Académique Documents
Professionnel Documents
Culture Documents
www.elsevier.com/locate/brainres
Research Report
ab st rac t
Article history:
Observers often fail to detect substantial changes in a visual scene. This so-called change
blindness is often taken as evidence that visual representations are sparse and volatile.
This notion rests on the assumption that the failure to detect a change implies that
Keywords:
representations of the changing objects are lost all together. However, recent evidence
EEG
suggests that under change blindness, object memory representations may be formed and
stored, but not retrieved. This study investigated the fate of object memory representations
when changes go unnoticed. Participants were presented with scenes consisting of real
Attention
world objects, one of which changed on each trial, while recording event-related potentials
Inattentional blindness
(ERPs). Participants were rst asked to localize where the change had occurred. In an
Awareness
additional recognition task, participants then discriminated old objects, either from the
Consciousness
pre-change or the post-change scene, from entirely new objects. Neural traces of object
memories were studied by comparing ERPs for old and novel objects. Participants
performed poorly in the detection task and often failed to recognize objects from the
scene, especially pre-change objects. However, a robust old/novel effect was observed in
the ERP, even when participants were change blind and did not recognize the old object.
This implicit memory trace was found both for pre-change and post-change objects. These
ndings suggest that object memories are stored even under change blindness. Thus,
visual representations may not be as sparse and volatile as previously thought. Rather,
change blindness may point to a failure to retrieve and use these representations for
change detection.
& 2013 Elsevier B.V. All rights reserved.
1.
Introduction
n
Correspondence address: Charit Universittsmedizin, Institut fr Medizinische Psychologie, Luisenstr. 57, 10117 Berlin, Germany.
Fax: +49 30 450 529 990.
E-mail address: niko.busch@charite.de
0006-8993/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.brainres.2013.05.014
108
2.
Results
2.1.
Behavioral results
Table 1 Behavioral performance in the change localization task and the old/new recognition task. The table gives average
proportion correct (and standard error of the mean) relative to the number of pre-change and post-change trials.
Pre-change
Rec. correct
Rec. incorrect
Post-change
Loc. correct
Loc. incorrect
Loc. correct
Loc. incorrect
0.30 (0.02)
0.17 (0.01)
0.28 (0.02)
0.24 (0.01)
0.46 (0.03)
0.03 (0.01)
0.30 (0.02)
0.21 (0.02)
109
2.2.
EEG results
Fig. 1 Schematic time course of a trial sequence. Each trial started with a display consisting of 9 nameable, familiar objects. On
each trial, one of these objects changed after a brief interruption of the display. Participants rst had to detect the change by
clicking on the location where the change had occurred. Following the detection response, two single objects were presented,
one of which was a novel object that had never been presented before in this experiment. The other object was either the prechange or the post-change object from the same trial. Participants reported which of the two objects had been presented before
on this trial. The ERP analysis focused on the difference between old and novel probe objects.
110
Fig. 3 Event related potentials (ERPs) for old and novel probes (averaged across participants) on trials with correct detection and
recognition. Left panel: difference between old and novel probe objects tested with a two-tailed t test. The old/novel effect was
tested at each sampling point and electrode, and the resulting t-values are color-coded. T-values corresponding to a p40:01 are
masked in grey. Each horizontal line represents one electrode, and electrodes are sorted according to their anteriorposterior
position. This analysis revealed temporally and spatially widespread old/new effects, starting at around 200 ms at frontal
electrodes, followed by a more parietal distribution from 500 to 700 ms, and a fronto-central distribution from 800 ms. Based on
this analysis, two spatio-temporal regions of interest were selected for further analyses of the old/novel effect. Middle panel: ERP
time courses averaged across electrodes within the central region of interest (see white markers in the topography), used for
testing the old/novel effect between 200 and 700 ms (grey shaded area). The topography shows the difference between old and
novel probe objects. Right panel: ERPs and difference topography for the old/novel effect between 800 and 1200 ms at right
fronto-central electrodes.
2.2.1.
ERPs 200700 ms
t19 3:11; p 0:006. In other words, even when the behavioral response indicated that the participant was not able to
discriminate an old from a novel probe object, the neural
response to old probe objects was nevertheless affected by
the prior presentation of this object.
A second ANOVA with factors probe novelty (old/novel) and
probe type (pre-change/post-change) compared the old/novel
effect under incorrect detection and recognition between prechange and post-change probes. This analysis conrmed the
main effect of probe novelty (F1;19 17:02; p 0:001; no interaction). Follow up tests conrmed that the old/novel difference was signicant for both pre-change probes (t19 3:74;
p 0:001; two-tailed t test) and post-change probes (t19
2:708; p 0:014).
2.2.2.
ERPs 8001200 ms
At longer latencies and frontocentral electrodes (Fig. 4, bottom panel), the polarity of the old/novel effect depended on
the recognition of the probe object, as indicated by a probe
novelty recognition interaction F1;19 30:97; po0:001: when
recognition of the probe object was correct, ERPs for novel
probe objects were more negative-going than ERPs for old
objects. The opposite pattern was found when recognition
was incorrect, ERPs for novel probe objects were more
positive than ERPs for old objects. Moreover, the old/novel
effect was strongest when both detection and recognition
were correct, as indicated by a probe novelty detection
recognition interaction F1;19 4:68; p 0:043. Post hoc test
111
Fig. 4 Event related potentials (ERPs) for old and novel probes averaged across participants. Upper panel: ERPs at central
channels (for channel locations see Fig. 3). The grey shaded area shows the old/novel effect in the time range from 200 to
700 ms. Statistically signicant old/novel effects were found in all conditions, even when detection and recognition had failed.
Bottom panel: ERPs at right fronto-central channels. The grey shaded area shows the old/novel effect in the time range from 800
to 1200 ms. The polarity of the late old/new effect depends on the success of recognition: ERPs are more positive for the object
which participants decided to report as old.
3.
Discussion
112
indicate the involvement of post-retrieval and possibly postdecisional monitoring and evaluation processes (Nessler
et al., 2001).
The present study found evidence for implicit object
memory traces under change blindness: a robust old/novel
ERP effect that was independent of change detection and
recognition of the changing object. The extant literature on
change blindness has been inconclusive as to whether implicit object representations under change blindness actually
exist. A number of behavioral studies have addressed this
question by asking whether observers can guess the location
of a change or whether their performance in a subsequent
task is modulated by the identity of the changing object, even
when they are not aware of any change. Some studies have
reported implicit effects on behavior in spite of change
blindness (Fernandez-Duque and Thornton, 2003, 2000;
Thornton and Fernandez-Duque, 2000; Laloyaux et al., 2006),
while others have argued that these effects can be explained
without implicit representations (Mitroff and Simons, 2002;
Mitroff et al., 2002). Likewise, EEG and fMRI studies have
provided only limited empirical support for implicit object
representations under change blindness, with some studies
nding neural signatures of implicit change detection
(Schankin and Wascher, 2008, 2007; Fernandez-Duque et al.,
2003), but with as many failures (Pourtois et al., 2006; Busch
et al., 2010b, 2010a; Koivisto and Revonsuo, 2003; Niedeggen
et al., 2001; Henderson and Orbach, 2006; Eimer and Mazza,
2005) or inconclusive ndings (Beck et al., 2001). It is important to note that most previous ERP studies that searched for
implicit or preserved object representations under change
blindness have looked at neural activity at the time the
change occurred (usually time-locked to the onset of the
changed display). These analyses have focused on the difference between trials without any change and change trials in
which subjects were blind to the change. However, comparing undetected changes to no-changes can only reveal implicit detection of change; this procedure is not suited for
revealing implicit object memory representations under
change blindness. Thus, this is the rst ERP study demonstrating electrophysiological correlates of implicit object
memory traces under change blindness by combining a
change detection paradigm with a subsequent recognition
memory task. This nding is in line with behavioral studies
that found preserved memory for objects even when a
change to that object had not been detected (Varakin and
Levin, 2006; Hollingworth, 2005; Hollingworth and Henderson,
2002; Beck et al., 2007; Beck and van Lamsweerde, 2011). It
should be noted that old items in the present study were, in
fact, always old and changed items. Thus, the old/novel ERP
effect shows that participants stored information about these
objects even under change blindness. However, this comparison does not reveal whether such memories include information about the object's status as changed or unchanged.
Future studies should compare old and changed items to
old, but unchanged items to clarify this question.
In sum, the ndings of the present study suggest that
visual representations may not be as sparse and volatile as
previously thought. Rather, change blindness may point to a
failure to retrieve and use these representations for change
detection. Thus, the change blindness phenomenon shows
113
that we see less than we believe, but we also see more than
we know.
4.
Experimental procedures
4.1.
Participants
4.2.
114
4.3.
v
u
u1 n
earlyi
latei 2
GDIearly;late t
n i 1 GFPearly GFPlate
where earlyi/latei are the average-referenced voltages at electrode i, and GFP refers to the global eld power, which is
equivalent to the standard deviation of the voltages across
electrodes (Murray et al., 2008). The GDI is an index of
conguration differences between two electric elds, independent of their strength. To test this measure of topographical difference, a permutation test was computed as
follows: (1) single-subject topographies were randomly reassigned to the two conditions, (2) the grand-average topography was recalculated, and (3) the GDI of this new grandaverage was calculated. This procedure was repeated 5 times.
This permutation tests the null hypothesis that all differences between the two topographies are not due to actual
condition effects, but rather due to random variance across
participants. A p-value was calculated by comparing the
empirical GDI to the distribution of GDI s obtained by the
permutation procedure.
Acknowledgments
This work was supported by the German Research Foundation (Grant # BU2400/1-1). I thank Irena oreva for help with
stimulus collection and data acquisition.
r e f e r e n c e s
Beck, D.M., Rees, G., Frith, C.D., Lavie, N., 2001. Neural correlates
of change detection and change blindness. Nat. Neurosci. 4 (6),
645650.
Beck, M.R., Levin, D.T., 2003. The role of representational volatility
in recognizing pre- and postchange objects. Percept.
Psychophys. 65 (3), 458468.
Beck, M.R., Peterson, M.S., Angelone, B.L., 2007. The roles of
encoding, retrieval, and awareness in change detection. Mem.
Cognit. 35 (4), 610620.
Beck, M.R., van Lamsweerde, A.E., 2011. Accessing long-term
memory representations during visual change detection.
Mem. Cognit. 39 (3), 433446.
Becker, M.W., Pashler, H., 2002. Volatile visual representations:
failing to detect changes in recently processed information.
Psychon. Bull. Rev. 9 (4), 744750.
Becker, M.W., Pashler, H., Anstis, S.M., 2000. The role of iconic
memory in change-detection tasks. Perception 29 (3), 273286.
Blackmore, S.J., Brelstaff, G., Nelson, K., Troscianko, T., 1995. Is
the richness of our visual world an illusion? Transsaccadic
memory for complex scenes. Perception 24 (9), 10751081.
Brainard, D.H., 1997. The psychophysics toolbox. Spat. Vis. 10 (4),
433436.
115