Letters Response
Attentional pointers: response to Melcher
Patrick Cavanagh1,2, Amelia R. Hunt3, Arash Afraz4 and Martin Rolfs5
1 Laboratoire Psychologie de la Perception, Universite´ Paris Descartes, 45 rue des Saints Pe`res, 75006 Paris, France
2Vision Sciences Laboratory, Harvard University, 33 Kirkland Street, Cambridge, MA 02138, USA
3School of Psychology, William Guild Building, University of Aberdeen, Aberdeen, AB24 2UB, UK
4McGovern Institute for Brain Research, Massachusetts Institute of Technology, 43 Vassar Street, Cambridge, MA 02139, USA
5Department of Psychology, New York University, 6 Washington Place, New York, NY 10003, USA
Melcher appears to have misunderstood our opinion piece
on attention pointers [1], or perhaps we did not state it
clearly enough. We did not claim that when it comes to
visual stability, the attention system does the work. The
oculomotor system does the work and the attention system
comes along for the ride. The performance benefits that
comprise the central properties of spatial attention appear
to be parasitic on the functions of the eyemovement control
system [2]. As reviewed by Awh et al. [3], stimulation of
cells in the saccade control centers triggers attentional
benefits at corresponding retinotopic locations. However,
when the eyes move, the activity in saccade centers is
shifted to the locations that targets of interest will have
following the saccade. Attention does not do this shift work,
but the consequence is that attentional benefits will then
be at the appropriate locations following the saccade.
To summarize, the saccade control centers maintain a
set of potential target locations as peaks of activity. If the
activity at one location crosses a movement threshold
(while subcortical neurons are controlling the fixation
pause [4]), a saccade is triggered. At lower levels of activity,
these location pointers confer attentional benefits at the
corresponding retinotopic locations in early visual cortices
[3]. That is why we called these peaks of activity in saccade
centers attentional pointers. Shifting these pointers at the
time of saccades is a service of the oculomotor system that
keeps the attention pointers appropriately aligned with
targets of interest.
By contrast, we were clear that there is nothing in these
pointers that relates to features. They are not, in our
proposal, feature-specific in any way, although experi-
ments might show them to be so in the future. We did
note that there must necessarily be a link between the
location information of a target (its attention pointer) and
some other set of identity information about the target. The
combination of the identity with location would correspond
to the putative structure of ‘object files’ [5]. The nature of
this link is central to the understanding of visual proces-
sing in general, as well as spatiotopy in particular, as
pointed out by Melcher [1].
We also suggested that these attention pointers, coupled
with a link to their target identity, allow high-level spa-
tiotopy: we know where the target is after the eye move-
ment so we know what its properties are even before we
start to re-encode them from its new location. This can
easily lead to spatiotopic priming (e.g. [6]). By contrast, it is
not clear that the information stored about an object
includes the adaptation state of the cells that are encoding
it. It is this information that would have to be transferred
to generate the spatiotopic aftereffects (e.g. [7]) that have
been difficult to replicate (e.g. [8]). Therefore our distinc-
tion between the presence of high-level spatiotopy (for
identity, in the form of priming) and the absence of low-
level spatiotopy (for aftereffects) is straightforward. The
link between location and identity is a very general re-
quirement of visual processing and the lack of evidence of
where or how it works is a challenge for all visual science
and is not a weakness of our proposal in particular.
Finally, these attention pointers and their shifts at the
time of saccades are sufficient to explain that apparent
motion seen between two successive stimuli is based on
spatial coordinates and not retinal coordinates when a
saccade intervenes between the first and second stimulus
[9]. However, Melcher points out that some types of appar-
ent motion cannot be explained by mere shifts between the
centers of the two stimuli. In particular, if the stimuli have
different shapes, a transformation is seen between the two
[10] even across a saccade [1]. Clearly, a shift of a pointer is
not sufficient to explain this phenomenon. Our interpreta-
tion is not that the pointers are linked to features such as
shape, but that the activity pattern on the saccadic map is
shaped like the target. We have evidence of this in a recent
remapping study using fMRI [11]. In line with previous
fMRI studies of remapping (e.g. [12]), saccades generated
BOLD activity in early retinotopic cortices at an expected
post-saccadic retinotopic location of a target even though
the target was removed before the saccade landed and so
was never present at that location. In our case, the target
had a wedge shape that changed orientation on each
saccade, and indeed the spatial pattern of the remapped
activity was correspondingly wedge shaped, rotating in
step with the stimulus.
Given that the activity pattern confers attentional ben-
efits on the corresponding location in early cortex (which is
where we were measuring the BOLD activity), this shaped
pattern of activity, if it holds up in future studies, also
makes sense of object-based attention [13] where attention
to one part of an object spreads throughout the object.
Admittedly, calling this target-shaped activity pattern a
pointer goes beyond the usual meaning of ‘pointer’. Never-
theless, it does ‘point’ in that it indexes all locations within
the object. A better label could emerge but ‘target-shaped
attention pointers’ is not it, so we still favor ‘attention
pointers’.Corresponding author: Cavanagh, P. (patrick.cavanagh@parisdescartes.fr).
Update Trends in Cognitive Sciences Vol.14 No.11
474

References
1 Melcher, D. (2010) The missing link for attention pointers: comment on
Cavanagh et al. Trends Cogn. Sci. 14, 473
2 Corbetta, M. and Shulman, G.L. (2002) Control of goal-directed
and stimulus-driven attention in the brain. Nat. Rev. Neurosci. 3,
201–215
3 Awh, E. et al. (2006) Visual and oculomotor selection: links, causes
and implications for spatial attention. Trends Cogn. Sci. 10, 124–
130
4 Cohen, B. and Henn, V. (1972) Unit activity in the pontine reticular
formation associated with eye movements. Brain Res. 46, 403–
410
5 Kahneman, D. et al. (1992) The reviewing of object files: object-specific
integration of information. Cogn. Psychol. 24, 175–219
6 Wittenberg, M. et al. (2008) Perceptual evidence for saccadic updating of
color stimuli. J. Vis. 8, 1–9
7 Melcher, D. (2005) Spatiotopic transfer of visual-form adaptation across
saccadic eye movements. Curr. Biol. 15, 1745–1748
8 Afraz, A. and Cavanagh, P. (2009) The gender-specific face aftereffect is
based in retinotopic not spatiotopic coordinates across several natural
image transformations. J. Vis. 9, 1–17
9 Rock, I. and Ebenholtz, S. (1962) Stroboscopic movement based on
change of phenomenal rather than retinal location. Am. J. Psychol.
75, 193–207
10 Tse, P. et al. (1998) The role of parsing in high-level motion processing.
In High Level Motion Processing (Watanabe, T., ed.), pp. 249–266, MIT
Press
11 Knapen, T. et al. (2010) Phase-encoded fMRI investigation of
retinotopic remapping responses. J. Vis. 10, 510
12 Merriam, E.P. et al. (2007) Remapping in human visual cortex.
J. Neurophysiol. 97, 1738–1755
13 Duncan, J. (1984) Selective attention and the organization of visual
information. J. Exp. Psychol. Gen. 113, 501–517
1364-6613/$ – see front matter ! 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tics.2010.08.006 November 2010, Vol. 14, No. 11
Book Review
Interpreting brain images: reflections on an
adolescent field
Foundational Issues in Human Brain Mapping
by Stephen Jose´ Hanson and Martin Bunzl, The MIT Press, 2010. $38.00/£28.95 (321 pp.) ISBN 978-0-262-51394-4
Nikolaus Kriegeskorte
Medical Research Council, Cognition and Brain Sciences Unit, Cambridge, CB2 7EF, UK
Functional brain imaging ismaturing, but
still adolescent. The field has developed a
rich toolbox of experimental and data an-
alytical techniques and is addressing an
ever expanding range of questions about
brain and mind – at various levels of
methodological rigor. Some of these ques-
tions (e.g. romantic love) are difficult to
pin down with science. Occasionally,
results are naively overinterpreted in sci-
entific papers and in the media. It is appropriate then to
reflect on our basic assumptions.
This edited book is a useful collection of conceptual and
methodological arguments on how to best use imaging to
learn about cognition and brain function. The issues range
from experimental design and analysis to theoretical in-
terpretation of the results, spanning multiple disciplines,
including statistics, computational modeling, cognitive
and brain theory, and philosophy.
Imaging seems to explain the fluff of the psyche at the
level of the hardware and it combines the prestige of
serious science with the broad appeal of intuitive images.
This combination is dangerously seductive. The brain blob
has the power to make us believe, however tenuous its link
to the proposition in question.
But brain images are not like photographs, direct and
simple reflections of their content matter. We must not
jump from a colored blobs to mental conclusions. Instead
we need to consider the intervening inferential steps: the
blob through the statistics reflects the imaging signal,
which reflects the hemodynamic response to neuronal
activity, which in turn might or might not underlie the
mental phenomenon (Roskies; parenthetical names refer
to chapter authors). These perils notwithstanding, our
intuition is fundamentally correct: brain images really
do afford discovery (‘Will any region be found?’ ‘If so, which
one?’) and substantial theoretical insight into brain infor-
mation processing.
Since the cognitive revolution, we have been construct-
ing theories about information processing in the brain.
Initially our models of cognition were based on behavioral
data alone. Despite ingenious methods for inferring inter-
nal processes, cognitive theory is vastly underconstrained
by behavioral data: there are many different theories
consistent with the data. Brain imaging can help not
only to localize functions anatomically, but also to better
constrain theories at the cognitive and neural levels
(Coltheart; alternative perspectives by Mole and Klein;
Harman; Loosemore and Harley; and Bechtel and
Richardson).
One challenge of engineering (or reverse engineering)
an information-processing system is functional decomposi-
tion: how is the complex process to be divided into func-
tional subcomponents implemented in separate physical
parts of the system?
In building computers and algorithms, we divide the
system into modules such that interactions across bound-
aries are limited. This enables us to reason about the
system at a higher level of description, where we can safely
Corresponding author: Kriegeskorte, N. (nikolaus.kriegeskorte@mrc-cbu.cam.
ac.uk).
Update Trends in Cognitive Sciences Vol.14 No.11
475