Neuron

Previews

Li, H., Fertuzinhos, S., Mohns, E., Hnasko, T.S.,Verhage, M., Edwards, R., Sestan, N., and Crair,M.C. (2013). Neuron 79, this issue, 970–986.

Morris, J.A., Royall, J.J., Bertagnolli, D., Boe, A.F.,Burnell, J.J., Byrnes, E.J., Copeland, C., Desta, T.,Fischer, S.R., Goldy, J., et al. (2010). Proc. Natl.Acad. Sci. USA 107, 19049–19054.

Nieto, M., Monuki, E.S., Tang, H., Imitola, J.,Haubst, N., Khoury, S.J., Cunningham, J., Gotz,M., and Walsh, C.A. (2004). J. Comp. Neurol.479, 168–180.

Pallas, S.L., Roe, A.W., and Sur, M. (1990).J. Comp. Neurol. 298, 50–68.

Neuron 79, S

Schaeren-Wiemers, N., Andre, E., Kapfhammer,J.P., and Becker-Andre, M. (1997). Eur. J. Neuro-sci. 9, 2687–2701.

Wagener, R.J., David, C., Zhao, S., Haas, C.A.,and Staiger, J.F. (2010). J. Neurosci. 30, 15700–15709.

Dopamine: Burning the Candle at Both Ends

John M. Pearson1,* and Michael L. Platt1,2,*1Duke Institute for Brain Sciences, Center for Cognitive Neuroscience, Department of Neurobiology2Departments of Biological Anthropology and Psychology and NeuroscienceDuke University, Durham, NC 27708, USA*Correspondence: pearson@neuro.duke.edu (J.M.P.), platt@neuro.duke.edu (M.L.P.)http://dx.doi.org/10.1016/j.neuron.2013.08.011

Dopamine neurons are well known for signaling reward-prediction errors. In this issue, Matsumoto andTakada (2013) show that some dopamine neurons also signal salient events during progression through avisual search task requiring working memory and sustained attention.

Imagine yourself on the hunt. This could

be the hunt for the last vegetarian option

at a department lunch or for a rare first

edition of Darwin’s ‘‘On the Expression

of Emotions in Man and Animals’’ at a

local flea market. Either way, the search

is on, and all of your senses are bent

toward that single goal. But what exactly

is it that drives you? What in your brain

is responsible for that sense ofmotivation,

a drive perhaps independent of your relish

at the attainment of the goal? What sets

your expectations, registers themismatch

between anticipation and experience,

and makes sure you don’t waste time on

a worthless search again? And what,

above all, is facilitating the laser-like

intensity with which your eyes—sifting,

sorting, homing in—scan the world

around you? The answer, of course, is

complicated. It is complicated because

it is biology. But there is also a simple

answer, one that comes up over and

over in studies of what drives us. That

answer is dopamine.

For more than a decade, dopamine has

been the darling of cognitive and systems

neuroscience. Synthesized by only a few

neurons (a mere 400,000) in the midbrain

but projected broadly across the telen-

cephalon, it has come to play an outsized

role in our thinking about learning, mem-

ory, movement, and motivation. This

stems in part from the key role it plays in

maladies such as Parkinson’s disease,

addiction, and schizophrenia, but also

from the emergence in the late 1990s of

highly influential computational theories

of its function (Berridge and Robinson,

1998; Schultz et al., 1997). Yet

despite the highly structured connectivity

patterns of midbrain dopamine neurons

(Haber and Knutson, 2010), most theories

have posited a single, unified role for their

function.

The last few years, however, have

witnessed a newwave of findings demon-

strating previously neglected diversity in

dopamine function, picking up on earlier

observations that dopaminergic cells

respond to salient events (Bromberg-

Martin et al., 2010; Horvitz, 2000; Matsu-

moto and Hikosaka, 2009; Redgrave and

Gurney, 2006) and perhaps even aversive

outcomes (Fiorillo, 2013; Horvitz, 2000;

Matsumoto and Hikosaka, 2009). These

findings raise the possibility that dopa-

mine release might subserve multiple

functions, conveying different signals to

different parts of the brain in order to

meet a variety of behavioral demands.

Yet a clear delineation of what functions

these disparate signals perform has

been lacking.

In this issue, Matsumoto and Takada

(2013) set out to remedy this gap by

studying the diversity of dopamine

signaling across the midbrain during

cognitive performance. To do this, they

recorded single neurons from the ventral

tegmental area (VTA) and substantia

nigra pars compacta (SNc) in monkeys

performing a visual search task for fluid

reward. On most trials, monkeys were

first shown a cue indicating whether a

large or small reward would be delivered

for a correct response. This cue was

followed by a sample stimulus (a slanted

line). The monkeys were then shown an

array of slanted lines (two, four, or six

items), among which they had to search

for a match to the sample stimulus.

Monkeys indicated a match by visually

fixating the matching target.

Previous work has shown that dopa-

mine is necessary for maintaining working

memory (Li and Mei, 1994; Sawaguchi

and Goldman-Rakic, 1991, 1994; Wata-

nabe et al., 1997; Williams and Goldman-

Rakic, 1995), as well as for facilitating

visual perception (Noudoost and Moore,

2011), and thus might be released in

response to the display of the target

eptember 4, 2013 ª2013 Elsevier Inc. 831

A C DB

Figure 1. Dopamine Neurons Respond to Events in a Working-Memory Task(A) Putative dopaminergic neurons in both SNc and VTA respond more strongly to cues predictinghigh-reward trials than to those predicting low-reward trials.(B) Cells in the dorsolateral SNc respond to the presentation of a visual stimulus when that stimulus mustbe maintained in working memory, but not when it is irrelevant to task performance.(C) Cells respond more strongly to the onset of smaller, easier search arrays than to larger, more difficultarrays. Responsive cells are more strongly concentrated in the medial SNc and VTA.(D) Cells respond more strongly to the location of a search target when that target is located in a larger(more difficult) array. More red indicates a stronger population response.All comparisons are relative within vertical columns.

Neuron

Previews

cue. Yet, this should only be necessary

when the information in the sample stim-

ulus is needed for the upcoming search.

To test this, the authors interleaved

blocks of the match-to-sample task with

blocks of a second visual search task. In

this second task, a slanted line stimulus

was again presented, but the search array

consisted of unrelated shapes (triangles

and squares). The monkey’s task was

then simply to locate the lone triangle,

which ‘‘popped out’’ from the array. For

this task, the initial stimulus was un-

necessary, and no working memory was

required.

The results of Matsumoto and

Takada’s experiment are summarized in

Figure 1. As expected, dopamine neurons

responded more strongly to the cue

advertising a large reward than to the

cue for a small reward (A). More impor-

tantly, cells responded much more

strongly to the sample stimulus when it

was needed for the upcoming search

than when it was irrelevant, suggesting

that dopamine release from midbrain

neurons contributes to the working mem-

ory requirements of the match-to-sample

task (B). In addition, dopamine cells fired

more strongly to the onset of smaller,

832 Neuron 79, September 4, 2013 ª2013 El

easier arrays than to larger, harder ones

(C) and responded more strongly when

monkeys found targets in large arrays

than in small ones (D). These results are

consistent with conventional motiva-

tion or reward prediction theories, which

predict a smaller dopamine release for a

lower probability of reward (large arrays)

and a larger dopamine release when a

low reward is actually obtained (large ar-

rays again). However, the target choice

signals Matsumoto and Takada observed

occurred after the monkeys fixated the

target but before delivery of the reward,

implying that these, too, encoded an

expectation of reward. In fact, the same

signals were present in trials where the

monkeys made incorrect choices, con-

sistent with the interpretation that they

reflected monkeys’ subjective expecta-

tions rather than the reward outcome or

a prediction error.

The authors’ most intriguing finding

resulted from an analysis of which neural

responses were present in which cells.

Although nearly all cells responded to

the onset of the reward cue, cells re-

sponding to the sample stimulus were

found almost entirely in dorsal and lateral

regions of the midbrain, probably within

sevier Inc.

the SNc. By contrast, cells responsive to

the size of the search array were more

concentrated in medial and ventral re-

gions, and there was a correlation

between effect size and recording depth,

most likely in the VTA. Such a gradient in

function is broadly consistent with known

anatomy: the SNc projects primarily to

dorsolateral sensorimotor structures,

whereas the VTA projects primarily to

medial and limbic cortical areas associ-

ated with learning and motivation (Haber

and Knutson, 2010). These observations

endorse the authors’ conclusion that

responses to the sample cue facilitate

working memory by releasing dopamine

in the dorsolateral prefrontal cortex.

They are likewise consistent with the

observation that factors influencing task

difficulty are processed preferentially by

systems responsible for calculating moti-

vation and reward anticipation.

In addition to these tantalizing findings,

the study also raises a number of impor-

tant questions. Because the authors

used spike waveforms to identify putative

dopaminergic cells and recorded only

firing-rate responses, they could not

verify the actual amount of dopamine

released in response to task events;

such verification could be provided by

techniques such as voltammetry, which

measures catecholamine release with

millisecond precision. Furthermore, the

difficulty of recording from small brain-

stem regions limited the number of cells

recorded—enough so to suggest a

gradient in function, perhaps, but the

findings will benefit from replication.

Finally, although both the location and

timing of cell firing in response to the

sample cue are consistent with the

hypothesis that subsequent dopamine

release facilitates working memory, future

studies will need to verify this causally,

perhaps by showing that selective activa-

tion or inactivation of lateral SNc neurons

has an effect on the performance of

working memory.

What is most exciting about the work

by Matsumoto and Takada is the finding

that dopamine signaling in the brain is

more heterogeneous and computa-

tionally specific than commonly thought.

Their work shows that what has long

been known anatomically is also true

functionally, and it challenges other

scientists to begin working out the means

Neuron

Previews

by which the brain orchestrates region-

specific dopamine signaling. Just as

importantly, the finding that dopamine

neuron responses track cognitive func-

tion could prove to be valuable for our

understanding of Parkinson’s disease, in

which dopaminergic medications used

for the control of motor symptoms are

sometimes accompanied by cognitive

side effects. Further work delineating

the separate cognitive, motor, and

learning signals in the SNc and VTA

might eventually lead to better treat-

ments that preferentially target dopa-

mine’s role in movement while sparing

patients’ cognitive abilities. Yet much re-

mains to be done. For a long while yet, it

appears, the tiny dopaminergic midbrain

will continue to demand a large body

of work.

REFERENCES

Berridge, K.C., and Robinson, T.E. (1998). BrainRes. Brain Res. Rev. 28, 309–369.

Bromberg-Martin, E.S., Matsumoto, M., andHikosaka, O. (2010). Neuron 68, 815–834.

Fiorillo, C.D. (2013). Science 341, 546–549.

Haber, S.N., and Knutson, B. (2010). Neuropsy-chopharmacology 35, 4–26.

Horvitz, J.C. (2000). Neuroscience 96, 651–656.

Li, B.-M., andMei, Z.-T. (1994). Behav. Neural Biol.62, 134–139.

Matsumoto, M., and Hikosaka, O. (2009). Nature459, 837–841.

Neuron 79, S

Matsumoto, M., and Takada, M. (2013). Neuron 79,this issue, 1011–1024.

Noudoost, B., and Moore, T. (2011). Nature 474,372–375.

Redgrave, P., and Gurney, K. (2006). Nat. Rev.Neurosci. 7, 967–975.

Sawaguchi, T., and Goldman-Rakic, P.S. (1991).Science 251, 947–950.

Sawaguchi, T., and Goldman-Rakic, P.S. (1994).J. Neurophysiol. 71, 515–528.

Schultz, W., Dayan, P., andMontague, P.R. (1997).Science 275, 1593–1599.

Watanabe, M., Kodama, T., and Hikosaka, K.(1997). J. Neurophysiol. 78, 2795–2798.

Williams, G.V., and Goldman-Rakic, P.S. (1995).Nature 376, 572–575.

The Cerebral Emporium of Benevolent Knowledge

Patrick J. Mineault1 and Christopher C. Pack1,*1Montreal Neurological Institute, McGill University, Montreal, QC H3A 2B4, Canada*Correspondence: christopher.pack@mcgill.cahttp://dx.doi.org/10.1016/j.neuron.2013.08.012

Visual objects tend to be found in predictable combinations (e.g., pens with paper). How does the brainrepresent these regularities? In this issue of Neuron, Stansbury et al. (2013) use fMRI to study the brain’srepresentation of visual scene categories.

In a 1942 essay, Jorge Luis Borges

discusses the categorization of animals,

purportedly found in a fictitious Chinese

encyclopedia named the ‘‘Celestial

Empire of Benevolent Knowledge’’

(Borges, 1942). Animals therein are

classified into 14 fanciful categories,

including, ‘‘fabulous ones,’’ ‘‘those that

have just broken the flower vase,’’ and

‘‘those that look like flies when viewed

from a distance.’’ Borges uses this

example to suggest that any attempt

to categorize the contents of nature is

‘‘arbitrary and full of conjectures.’’

Nevertheless (again quoting Borges),

‘‘the impossibility of penetrating the

divine scheme of the universe cannot

dissuade us from outlining human

schemes, even though we are aware

that they are provisional.’’ In fact, such

schemes can be quite useful in sensory

neuroscience. A decade after Borges’s

essay, Barlow (1953) discovered neurons

that respond selectively to stimuli that

look like flies when viewed from a dis-

tance. These ‘‘fly detectors’’ were found

in the retinas of frogs and, hence, were

linked to a specific category of behavior

(feeding). Subsequently, Hubel and

Wiesel (1962) identified visual cortical

cells that were described as ‘‘simple’’

and ‘‘complex,’’ and these turned out to

be useful labels for understanding many

aspects of the visual cortex from anatomy

to computation.

More recent imaging studies have led

to the suggestion that neurons with

particular stimulus selectivities are clus-

tered together, forming brain modules

responsible for encoding rather abstract

categories of stimuli, including faces

(Tsao et al., 2006), places (Epstein and

Kanwisher, 1998), and buildings (Hasson

et al., 2003). Of course, the number of

such categories must be far greater than

the number of brain regions, which leads

to the profound question of how the brain

organizes such a vast quantity of visual

experience. In this issue of Neuron,

Stansbury et al. (2013) address this

question.

Stansbury et al. (2013) used fMRI

imaging of human subjects to study the

brain’s representation of visual scene

categories, defined as classes of images

that contain similar co-occurrences of

individual objects. For example, a scene

that contains a building and a car is

more likely to belong to the category

‘‘cityscape’’ than to the category

‘‘nautical.’’ Obviously, one object (e.g.,

a tree) can be found in more than one

scene (e.g., cityscape and rural), and

eptember 4, 2013 ª2013 Elsevier Inc. 833