supporting online material for - science...supporting online material for neural responses to...

www.sciencemag.org/cgi/content/full/316/5831/1622/DC1

Supporting Online Material for

Neural Responses to Taxation and Voluntary Giving Reveal Motives for Charitable Donations

William T. Harbaugh,* Ulrich Mayr,* Daniel R. Burghart

*To whom correspondence should be addressed. E-mail: [email protected] (U.M.) or [email protected] (W.T.H.)

Published 15 June 2007, Science 316, 1622 (2005)

DOI: 10.1126/science.1140738

This PDF file includes

Materials and Methods Fig. S1 Tables S1 to S7 References Protocol

Supporting Online Material for

Neural Responses to Taxation and Voluntary Giving Reveal Motives for

Charitable Donations

William T. Harbaugh, Ulrich Mayr, and Daniel R. Burghart

1 Design, Procedure, and Sample

Subjects were scanned while they were presented with transfers that affected their own

account (starting amount $100) and the account of a local charity. Half the transfers were

mandatory, to resemble taxation, while the other half were voluntary. We explained that

the experimenters would not know their choices, and that one mandatory and one voluntary

transfer would be randomly chosen and implemented after the experiment. Events for each

trial occurred as presented in the time line shown in Figure 1. After a 1 second fixation dot,

the screen revealed whether this trial’s transfer was mandatory or voluntary as well as the

dollar amount change to the accounts of the subject and the charity. After 9 seconds, two

vertically aligned labels were added in the lower portion of the screen, specifying the options

for vertically aligned buttons on a response box. For mandatory transfers, one of the labels

read “acknowledge” and the other “invalid key”. For voluntary transfers, one of the labels

read “accept” and the other “reject”. Label positions varied randomly from trial to trial.

Immediately after the subjects response, a 4-point satisfaction rating scale was shown, to

which subjects responded by pressing one of four laterally oriented keys on a button box.

The rating scale disappeared after 6 sec and there was a blank screen for an inter-trial period

that was randomly jittered between 6, 7, or 8 sec.

1

Our experimental design included variations in the amounts going to the charity and the

amounts going to the subject (Fig. 1b). The main purpose for these manipulations was to

provide sufficient variation in the “price of giving” to elicit a range of individual responses

and to reduce subject fatigue.1 Each subject participated in three 13 minute runs of 38

trials. Within each run, each cell in the design was presented twice: once as a mandatory

transfer and once as a voluntary transfer that could be accepted or rejected. The order of

these transfers was random both within- and between-subjects.

Our sample included 20 female subjects recruited from the population of economics and

psychology students at the University of Oregon. Female subjects were selected to control for

any potential inhomogeneities in brain structures between genders, and because variation in

charitable giving decisions was a critical component of our analysis and behavioral evidence

suggests greater price sensitivity to giving in females (S1).

All subjects provided informed consent, consistent with IRB guidelines, and were free of

any neurological or psychological disorder. Hardware problems resulted in lost behavioral

data for all three runs for one subject and scanning data for one run of another participant

which leaves us with 56 individual-runs from 19 subjects for analysis.

1.1 Scanning

All scanning occured with the same Siemens Allegra 3.0 Tesla machine. Functional scanning

parameters were an EPI sequence of 32 axial slices (4 mm, FOV=200; Voxel size= 3.125 x

3.125 x 4 mm; TR=2000 ms, TE=30 ms, Flip Angle = 80 degrees). Structural scans occurred

with an MPrage 3D volume acquisition (FOV=256, 160 slices 1mm thick, Matrix=256x256,

1mm x 1mm x 1mm resolution, TR=2500ms, TE=4.38ms, TI=1100, flip angle=8 degrees).

1The term “price of giving” refers to how much money a subject must sacrifice to give one dollar to thecharity.

2

2 GLM Analysis

2.1 Data Conversion

Funtional and anatomical databases were converted from DICOM to SPM Analyze format

using the LCNI’s MRIConvert program (http://lcni.uoregon.edu/downloads.html).

2.2 Pre-Processing

Pre-processing occurred with utilities from the FMRIB Software Library (FSL) 3.2 (available

at http://www.fmrib.ox.ac.uk/fsl). Four-dimensional functional databases were constructed

using the avwmerge utility. Functional data were also subjected to (1) the brain extrac-

tion tool (BET), (2) motion correction (MCFLIRT), (3) 5mm full-width half-max (FWHM)

spatial smoothing, and (4) high-pass temporal filtering at 9 cycles ( .025 Hz).

2.3 First Level Analysis

Individual-run data were modeled voxel-by-voxel using the following specification:

BOLDt = γ0 + γ1x1t + · · ·+ γ38x38t + εt

where BOLDt is the BOLD signal (post pre-processing) and xjt is an indicator variable

for experimental treatment j at time t convolved to the canonical haemodynamic response

gamma function with a sdev of 3 and a mean lag of 6s (i.e. xjt represents a convolved delta

function).

Events were modeled with a 9s duration beginning with the appearance of the screen

which specified if the trial was mandatory or voluntary and gave the dollar amounts of the

transfer. Because the response phase occurred at 9s, this means we did not use data involving

the motor response in computing the parameters for the COPEs.

3

We assumed that the error term, εt, was Gaussian following a first order autoregressive

(AR(1)) process. The Cochrane-Orcutt estimator was used in generating the weighting

matrix since it retains the first observation in the time series, despite first differences. This

is also called “pre-whitening” and corresponds to FMRIB’s improved linear model, or FILM.

2.4 Second and Third Level Analyses

Before higher level processing, parameter estimates from first level analyses were co-registered

to high-resolution (T1-weighted) images and affine transformed to standard MNI-152 space

using 12 degrees of freedom. Second and third level analyses proceed hierarchically com-

bining individual model parameters across runs (second level) and then across individuals

(third level). This hierarchical modeling occurs with FMRIB’s local analysis of mixed effects

(FLAME). See (S2) for a discussion of the modeling, and (S3) for a discussion of the simula-

tion methods associated with generating the efficient standard errors. Parameter estimates

from the third level analysis were compared with contrasts and the contrast statistics were

first tested against a z-threshold of 2.3 and then against a cluster threshold of p = 0.05.

3 Behavioral Analysis

3.1 Choices

Subject’s choices (in the voluntary conditions) were in line with predictions from economic

theory: as the price of giving to the charity increased, there was a lower likelihood of a transfer

being accepted. Figure 2A (in the text) demonstrates this result for differing levels of loss

of funds to the participant. It also shows that, conditioning on the price, participants were

more likely to accept transfers when the amounts at stake were smaller. To formally examine

the relationships between treatment variables and individual choices we estimated probit

4

regressions which explain the probability of accepting a voluntary transfer as a function of

payment to the subject and payment to the charity, as independent predictors. Two models

were run, one with random effects and the other with standard errors clustered by individual.

As can be seen in Table S1, both predictors were highly significant.

3.2 Subjective Ratings of Satisfaction

We analyzed the subjective satisfaction ratings (shown in Figure 2B) in two ways. First

as a function of treatment variables: money to subject, money to charity, and a predictor

representing the mandatory-voluntary contrast (Model 1 in Table S2). Second, we incorpo-

rated subject decisions in the payoff predictors, referred to in Table S2 as “Actual Payoffs”.

Subjective ratings of satisfaction were higher both when the subject received more money

and when the charity received more money. Also, subjective ratings were higher (on average)

in the voluntary treatments. However, the reduction of the mandatory-voluntary coefficient

when using the actual payoffs indicates that increases in subjective satisfaction in volun-

tary conditions is, in part, due to changes in subject payoffs. There is an almost reliable

trend for increased satisfaction as a function of the mandatory-voluntary contrast, even after

controlling for actual payoffs.2

4 ROI Analysis

4.1 Data Strategy

We generated anatomical regions of interest (ROIs) using masks from the International

Consortium of Brain Masks (ICBM) and the Talairach atlas as guides. ROI centroid coor-

dinates and volumes are in Table S3. To get functional ROIs we intersected the anatomical

2If we focus our analysis on those design cells where costly tradeoffs are made (i.e. the blue cells in Figure1B), the coefficient on “Voluntary” is highly reliable (b=0.509, p¡0.01).

5

ROIs with voxels showing significant activation in contrasts comparing (1) voluntary and

mandatory treatments, (2) payment amounts to the subject in mandatory treatments, (3)

payment amounts to the charity in mandatory treaments, and (4) decision difficulty. More

precisely, let V OL ≡ 1 ∪ 2 ∪ 3 ∪ 4, where the numbers correspond to the sets of voxels

showing significant activation in the above, respectively numbered contrasts. We then de-

fine ROIFunc. ≡ ROIAnat. ∩ V OL. In each of these functional ROIs, the (spatial) average,

individual-run BOLD signals were extracted using the featquery utility and read into Stata

9.0, a statistical software package.

We denote the raw BOLD signals (more precisely, we queried the post pre-processing

databases for each individual-run) from a (functional) ROI as BOLDROIt (for ease of ex-

position, we suppress individual-run subscripting). Observations were collected every other

second (i.e. TR=2) and we linearly interpolated to generate the “missing” observations. We

then calculate the percent change in the BOLD signal as

boldROIjt ≡

BOLDROIjt −BOLDROI

Baselinej

BOLDROIBaselinej

∗ 100, j = 1, . . . , 38

where we define the baseline on a treatment by treatment basis; it is the time averaged

activation in the three seconds preceding the onset of the stimulus. That is, in treatment j,

the baseline for an ROI is calculated as

BOLDROIBaselinej

≡ 1

3

−1∑τj=−3

BOLDROIτj

where τj = 0 corresponds to the onset of the stimulus in experimental treatment j.

6

4.2 Neural Evidence for Pure Altruism

Next we analyzed these ROIs by examining how activation in mandatory conditions covaries

with the experimental treatments. We estimated OLS specifications of the following form:

boldROI

ij = β0 + β1 ∗ Subjectij + β2 ∗ Charityij + εij

where boldROI

ij ≡∑16

τij=3 boldROIijτ which we refer to as activation or activity in a ROI. We

included a random effects component and clustered the standard errors by individual which

makes this equivalent to a hierarchical model (S4).3 Table S4 summarizes the results from

these regressions.

We found that activation in ventral striatal regions responded both to how much money

the charity received and how much money the subject received. So even when a transfer is

mandatory, we see higher activation in the nucleus accumbens (bilateral) and the caudate

(right only) when either the charity or when the subject receives more money.

4.3 Higher Activation in Voluntary Conditions

Next we examine differences in activation between voluntary and mandatory conditions. Our

estimating specification is

boldROI

ij = β0 + β3 ∗ V oluntaryij + εij

where V oluntaryij is an indicator variable equal to unity when condition j is voluntary, and

zero otherwise. Again we included a random effects component and clustered the standard

errors by individual. Table S5 shows regression results when we examined activation differ-

3Note that in this, and subsequent estimates, our findings are robust to the selection of (1) individualfixed-effects, (2) individual-run fixed or random effects, (3) clustering standard errors by individual-run, and(4) White corrected standard errors.

7

ences between the voluntary and mandatory conditions: Activation in the caudate (bilateral),

nucleus accumbens (right) and the insulae were all significantly higher. This suggests that

there is an additional hedonic benefit that is related to the ability to choose.

4.4 Neural Evidence for Warm Glow

The combination of the mandatory and voluntary conditions allows us to compare neural

responses when transfers are mandatory (akin to a tax-based transfer of funds) and when

they are executed by the participant in an act of free choice, as in charitable giving. However,

in our experiment not all is held constant between these mandatory and voluntary conditions:

Subjects can alter their payoffs in voluntary conditions when they choose to reject a transfer.

To control for the payoff differences, we included the variables Subject and Charity (which

incorporate subject decisions to accept or reject a voluntary transfer), in addition to the

indicator variable V oluntary, in our regressions. Thus, our estimating specification is given

by

boldROI

ij = β0 + β1 ∗ Subjectij + β2 ∗ Charityij + β3 ∗ V oluntaryij + εij

and we included a random effect component and clustered the standard errors by individual.

Table S6 summarizes the results from these estimations.

We found that activation in the caudate (bilateral) and the right nucleus accumbens

was higher in the voluntary treatments, even when controlling for payoffs. This provides

neural support to the warm glow hypothesis: there is larger activation in reward-related

areas when executing a charitable transfer, over and above what occurs in an analogous

mandatory transfer, even after controlling for the payoffs associated with subject choices.4

4If we estimate the above model using stakes, instead of payoffs, we also find higher activation in the ROIsin voluntary treatments. This provides neural support to the economic principle that removing a constrainton the set of alternatives cannot make decision makers worse off.

8

4.5 Predicting Choices with Mandatory Activation

Economic theory predicts that individuals will choose to give to the charity when the

marginal utility from doing so is greater than the marginal utility of keeping the money

for themselves. This can occur because individuals greatly value money going to the charity,

or because individuals do not value money for themselves. If we can think of marginal utility

as a representation of hedonic experience in the human brain, we can test this prediction by

using activation in mandatory conditions as explanatory variables for choices in voluntary

conditions (while still controlling for variation in activation due to the experimental treat-

ments). We restrict our observations to the nine voluntary conditions in which an explicit

tradeoff is made (i.e. the blue portion of Figure 2). We estimated probit regressions of the

following form:

Pr(Acceptij = 1) = β0 + β1 ∗ Subjectij + β2 ∗ Charityij + β3 ∗ boldROI

i∗ + εij

where Acceptij = 1 indicates that a subject(-run) accepted a proposed transfer in experimen-

tal treatment j, and boldROI

i∗ denotes activation in an ROI in a set of mandatory treatments.

More precisely, boldROI

i∗ ≡∑

j∈R boldROI

ij , where R is a subset of experimental treatments.

For the estimates denoted with an -S, R is the subset of mandatory treatments in which

money accrues only to the subject (the orange vertical section in Figure 2B). For those esti-

mates denoted with a -C, R is the subset of mandatory treatments in which money accrues

only to the charity (the green horizontal section in Figure 2B). And again we model a ran-

dom effects component and cluster the standard errors by individual. Results from these

estimations are tabulated in Table S7.

These estimates match with the theoretical predictions: if a person had higher activation

when they received money in mandatory treatments (i.e. their neural activity suggests that

they get a large reward from money), they are less likely to give money to the charity. Addi-

9

tionally, subjects who showed increased ventral striatum and insulae activation in mandatory

treatments when only the charity received money were more likely to give. This table also

shows that the results in Figure 4 are remarkably robust: activation in many areas that have

been identified with rewards reliably predict the choices individuals make.

10

5 Figures

.

Figure S1Timing of events in the fMRI during each trial

11

6 Tables

Table S1Predicting Choices with Payments to Subject

and Charity: Coefficient from Probit and Random Effects Probit

Predictor (1) (2)

$ to Subject 0.066∗∗∗ 0.130∗∗∗

(6.41) (10.79)

$ to Charity 0.046∗∗∗ 0.094∗∗∗

(5.46) (9.30)

Random eff. No YesSEs clustered Yes NoLog L -243.1 -146.8

Notes: n=504. Absolute value of z-statsin parenthesis. Constant not shown.∗∗∗denotes significance at the 1% level.See Section 3.1 for discussion.

12

Table S2Subjective Satisfaction Ratings as a Function of (1) Design Factors

and (2) Design Factors Incorporating Actual Payoffs

Predictor (1) (2)

Voluntary 0.422∗∗∗ 0.267∗

(4.53) (1.84)Transfer Amounts$ to Subject 0.042∗∗∗ -

(7.17)$ to Charity 0.046∗∗∗ -

(9.75)Actual Payoffs$ to Subject - 0.048∗∗∗

(7.41)$ to Charity - 0.065∗∗∗

(11.76)ThresholdsThresh 1 -0.998∗∗∗ -0.712∗∗∗

(5.82) (5.05)Thresh 2 0.262∗∗ 0.529∗∗∗

(2.42) (4.90)Thresh 3 1.273∗∗∗ 1.582∗∗∗

(8.97) (10.85)Log L -2191.4 -2158.7

Notes: n=2128. Absolute value of z-stats inparenthesis. Standard errors clustered by 19subjects. ∗∗∗denotes significance at the 1% level,∗∗at the 5% level, ∗at the 10% level.See Section 3.2 for discussion.

13

Table S3Regions of Interest

Location and Volume in MNI-152 Space

ROI x1 x2 x3 Anatomical Functional(mm) (mm) (mm) (mm3) (mm3)

Caudate (L) -8 4 4 1768 1720Caudate (R) 8 4 4 1768 1344

Insula (L) -34 18 -12 4168 3560Insula (R) 34 18 -12 4168 2096

NAcc (L) -10 10 -6 984 728NAcc (R) 10 10 -6 984 560

Notes: Coordinates in MNI-152 space. The coordinates listed (x1, x2, x3)are the distance of the ROI centroid from the origin, in millimeters (mm).See Section 4.1 for definitions of anatomical and functional ROIs

Table S4Activations in Six ROIs During Mandatory Conditions as a Function of

Transfer Amounts to Subject and Charity

Predictor Caudate (L) Caudate (R) NAcc (L) NAcc (R) Insula (L) Insula (R)

$ to Subject 0.00094∗ 0.00147∗∗ 0.00118∗∗ 0.00141∗∗ -0.00001 0.00039(1.68) (2.41) (2.10) (2.13) (0.02) (0.92)

$ to Charity 0.00243∗∗∗ 0.00267∗∗∗ 0.00191∗∗ 0.00288∗∗∗ 0.00033 0.00084(3.04) (3.07) (2.37) (3.03) (0.43) (1.38)

Adjusted R2 0.0058 0.0075 0.0040 0.0072 0.0000 0.0000

Notes: n=1064. Constant not shown. Absolute value of z-stats in parenthesis. Standard errorsclustered by 19 subjects. ∗∗∗denotes significance at the 1% level, ∗∗at the 5% level, ∗at the 10% level.See Section 4.2 for discussion.

14

Table S5Activation in Six ROIs as a Function of Mandatory-Voluntary Contrast


Voluntary 0.05992∗∗ 0.07019∗∗∗ 0.03566 0.08178∗∗ 0.04017∗ 0.03308∗

(2.43) (2.87) (1.14) (2.57) (1.86) (1.78)Adjusted R2 0.0020 0.0030 0.0000 0.0025 0.0010 0.0010


Table S6Activation in Six ROIs as a Function of Mandatory-Voluntary Contrast

and Design Factors Incorporating Actual Payoffs


Voluntary 0.05875∗∗ 0.06489∗∗ 0.02049 0.06777∗∗ 0.03680 0.02876(2.27) (2.53) (0.62) (2.03) (1.62) (1.48)

$ to Subject 0.00079 0.00110∗ 0.00201∗∗ 0.00211∗∗ 0.00039 0.00059(1.22) (1.73) (2.46) (2.54) (0.68) (1.22)

$ to Charity 0.00136∗ 0.00137∗ 0.00170∗ 0.00205∗∗ 0.00026 0.00052(1.72) (1.75) (1.69) (2.01) (0.37) (0.88)

Adjusted R2 0.0029 0.0041 0.0021 0.0049 0.0004 0.0010


.

15

Table S7Choices as a Function of Neural Responses From Pure Gains to

Charity and From Pure Gains to Subject for Each ROI

Predictor (1) (2) (3) (4) (5) (6)

$ to Subject 0.134∗∗∗ 0.135∗∗∗ 0.135∗∗∗ 0.133∗∗∗ 0.135∗∗∗ 0.133∗∗∗

(10.72) (10.70) (10.76) (11.05) (10.69) (10.94)

$ to Charity 0.094∗∗∗ 0.094∗∗∗ 0.094∗∗∗ 0.093∗∗∗ 0.094∗∗∗ 0.093∗∗∗

(9.24) (9.23) (9.26) (9.42) (9.24) (9.36)

Caudate-S (L) -6.103∗∗∗

(2.58)Caudate-C (L) 5.213∗∗∗

(2.59)Caudate-S (R) -3.803

(1.22)Caudate-C (R) 3.098∗

(1.69)NAcc-S (L) -7.284∗

(1.72)NAcc-C (L) 3.233

(1.17)NAcc-S (R) -8.412∗∗

(2.49)NAcc-C (R) 3.650∗

(1.93)Insula-S (L) -4.545∗∗

(2.01)Insula-C (L) 3.880∗∗

(2.32)Insula-S (R) -6.675∗∗∗

(2.86)Insula-C (R) 5.147∗∗

(2.17)Log-L -144.1 -145.6 -145.7 -144.7 -143.0 -144.2

Notes: n=504. Constant not shown. Absolute value of z-stats in parenthesis.∗∗∗denotes significance at the 1% level, ∗∗at the 5% level, ∗at the 10% level.-S denotes activation from mandatory conditions where only the subject receives money.-C denotes activation from mandatory conditions where only the charity receives money.See Section 4.5 for discussion.

16

.

References

S1. Andreoni, J. & Miller, J.H. Econometrica 70 (2), 737-753 (2002).

S2. C. F. Beckmann, S. Smith, M. Jenkinson, General Multi-Level Linear Modelling forGroup Analysis in FMRI (Tech. Rep. TR01CB1, Oxford University, 2001).

S3. M.W. Woolrich, T.E.J. Behrens, C.F. Beckmann, M. Jenkinson and S.M. Smith Multi-Level Linear Modelling for FMRI Group Analysis Using Bayesian Inference (Tech. Rep.TR03MW1, Oxford University, 2003).

S4. J. Wooldridge, Econometric Analysis of Cross Section and Panel Data. (MIT Press,Cambridge, MA, 2002).

17

18

Protocol:

Welcome:

This is an experiment about decision-making. It should take about 2 hours. All your decisions will be secret, and at the end you will be paid your earnings privately and in cash. A research foundation has provided the funds for this experiment.

We consider ourselves bound to keep the promises we are making to you in this protocol. We are not allowed to use any deception in this experiment - we will do everything that we say, and there will not be any surprises or tricks. We will start with a brief introduction, then go through consent forms with you, and then explain the experiment in more detail. You will make your decisions inside the brain imaging or “fMRI” machine. You start the experiment with $100 in your account. We will also start an account that is earmarked for the non-profit charity “Food for Lane County.” The initial balance for their account is $0. We will show you some information about Food for Lane County shortly, in case you are unfamiliar with their work. During the experiment, you will be shown a series of transfers of money between your account and the Food for Lane County account. Usually, though not always, these transfers involve money going from your account to the Food for Lane County account. The amounts involved in a single transfer range from $0 up to $45. Sometimes you will be allowed to decide whether you want to make these transfers. We’ll call these voluntary transfers. Sometimes the decisions will be made regardless of what you want, and those will be called mandatory transfers.

When the experiment is over, the computer will randomly pick one voluntary transfer and one mandatory transfer. You will be paid your starting balance of $100, adjusted by the amounts of these two transfers. Food for Lane County, will receive their starting balance of $0, also adjusted by the amount of these two transfers. You will be given the final balance of money in your account, privately and in cash. We will send Food for Lane County a check for the final balance of their account. The transfer amounts are set so that you can’t end up with less than $10 in your account. (You also get the $15 per hour payment.) Food for Lane County cannot end up with less than $0.

19

Food for Lane County:

This information is taken directly from the Food for Lane County website, http://www.foodforlanecounty.org Food for Lane County collects and distributes food to low-income residents of Eugene and Lane County. Please take a minute to read through this information, and let me know when you are done.

FOOD for Lane County's mission

The mission of FOOD for Lane County is to eliminate hunger by creating access to food. We

accomplish this by soliciting, collecting, rescuing, growing, preparing and packaging food for

distribution through a network of social service agencies and programs; and through public

awareness, education and community advocacy.

FOOD for Lane County is the regional food bank serving all of Lane County, Oregon. As the second

largest food bank in the state, FOOD for Lane County finds creative solutions to hunger and its root

causes. We believe a responsive food bank includes programs that help people help themselves. Food

banking also requires the participation of the whole community.

Programs and Services

FOOD for Lane County’s hunger relief efforts are focused on providing emergency food assistance and increasing self-sufficiency for our neighbors living on limited resources. We serve more than one in

five Lane County residents through emergency food assistance, nutrition education, advocacy and

self-sufficiency efforts.

Our mission of eliminating hunger by creating access to food is accomplished with tremendous support

from our community and through a variety of innovative programs. From Healthy Futures, which

teaches the basics of healthy eating and food shopping on a limited budget, to Food Rescue Express,

which recovers over 425,000 pounds of food from local restaurants, dormitories, hospitals and hotels

every year, we are working to build a stronger, healthier Lane County.

FFLC receives highest charity rating

FOOD for Lane County has received the highest charity rating (4 stars) from

Charity Navigator, a nonprofit organization that works to help charitable givers

make intelligent giving decisions. Charity Navigator provides information on more

than 3,000 charities and evaluates the financial health of each.

20

Secrecy, Anonymity, and Payments:

All the decisions that you make in this experiment are secret and anonymous. We will never be able to match your decisions to your name. The people in the brain imaging lab will not know what decisions you are making. The person who pays you your money and mails the check to Food for Lane County will not know your decisions. The rest of this section explains the careful procedures we use to make sure of this.

Here is a card with a code number on it. When you go to the brain imaging lab where you make decisions, you will be identified only by this number, not by your name. The people running the experiment will know when you are shown information and when you make a decision on the computer, but they cannot see what decisions you make. We will not keep any record of whose name goes with what code number. We have a list of the names of people who sign up for the experiment, but this is for scheduling purposes and will be destroyed when we are finished running these experiments. Furthermore, we will remove the code numbers from our data immediately after it is collected, so that no one will be able to determine your decisions even if they knew your code number. When you are finished with the brain imaging part of the study, the brain imaging staff will copy the file with your decision onto a USB flash drive like this. You will bring the drive back to this room, and we will use the following random procedure to pick which mandatory and which voluntary transfer counts for payment. You will sit down at a computer and insert the flash drive. No one else will be able to see the screen. All the mandatory transfers will be shown on the computer in a list, ordered just as you saw them in the experiment. An arrow will move rapidly up and down the list, and you will be able to stop the arrow by pushing a key. The transfer where the arrow stops is the one that counts for payment. The arrow moves too quickly for you to control its precise stopping point. This is just a visual way of picking a transfer randomly. After a mandatory transfer has been randomly picked, you will use the same procedure to pick a voluntary transfer. These two transfers will then be shown to you on the screen. Your initial $100 will be adjusted by the amount of the mandatory transfer. If you accepted the voluntary transfer, your payoff will be adjusted by that amount as well. Food for Lane County will receive their initial $0, plus any amount resulting from the transfers. The last screen on the computer will show these final amounts. You will show this to me, and I will give you your payment in cash, and I will also fill out a check to Food for Lane County. If you want to, you can go with the me to the mailbox and see me mail this check to Food for Lane County, so that you know that they will actually receive the money. So, I will know the final amounts in the accounts, but I won’t know what transfers the computer randomly picked, or your decisions.

21

You will need to fill out and sign a receipt for your payment. Your signature does not need to be legible. These receipts will be given to the UO finance office. The people handling these receipts do not know anything about the decisions that can be made in this experiment. This experiment will produce valid results only if the participants believe that their decisions are anonymous and secret. This is why we take these issues extremely seriously!

22

Experiment:

Here is how the experiment will work. In the actual experiment, you will be inside the fMRI scanner, and all information will be displayed on a computer monitor. You will be shown computer screens which specify an amount of money for you and an amount of money for Food for Lane County. We’ll call this a transfer. The amounts of money shown can be positive or negative, and they will either be added to or subtracted from the initial $100 that you have been given. Most, but not all, of the transfers involve money moving from your account to the Food for Lane County account. The food for Lane County account starts with $0 in it. Transfers are either voluntary or mandatory. A voluntary transfer means that you choose whether to allocate the money according to what is shown on the screen, or leave the amounts unchanged. A mandatory transfer means that you do not choose – the money gets transferred according to the numbers on the screen, whether you want that or not. Before each transfer you are told the amounts, and whether it is voluntary or mandatory. During the course of the experiment, you will see about 100 transfers. About half will be voluntary, half will be mandatory, and they will be in random order. Remember, at the end of the experiment, one mandatory and one voluntary transfer will be picked, at random. Since some transfers may be shown more than once, you might have made different responses to the same transfers. Our procedure will randomly draw one transfer from all of the voluntary transfers you have been shown, and one from all the mandatory transfers.

For the voluntary transfer, the computer will then check your choice and implement the option you chose. If it’s a mandatory transfer, it will be implemented automatically. Since you don’t know which transfer will be randomly picked, you should treat each as if it’s the one that counts. Transfers do not always involve you giving up one dollar for every dollar that goes to the charity. Sometimes the money taken from your account is increased before going to the charity, like when donations are matched by someone else. So, it might cost you $15 for you to give the charity $45. Sometimes, the money taken from your account is reduced before going to the charity. So it might cost $30 for you to give the charity $15. This is as if you give up something that is very valuable for yourself, but which is not quite as valuable for the charity. Sometimes, the charity simply receives money without it costing you anything. For example, the charity might gain $30 and your account balance would not change. This is as if someone else gave the money to the charity. Sometimes, money is simply taken from or put into your account, without affecting how much the charity gets. Finally, there are a few cases were nothing changes. Neither account balance changes. We will now show you a few examples and give you a short quiz, to make sure you understand the experiment.

23

Example 1: You might see this screen for 9 seconds:

The following transfer is

MANDATORY.

Change to

your account:

Change to Food for Lane County

account:

- $30

+ $30

A bit later, a message will be added to the lower screen half:


MANDATORY.

Change to

your account:


account:

- $30

+ $30

Acknowledge

Invalid Button

The buttons at the bottom will randomly switch locations. You will have 6 seconds to click the acknowledge button. After clicking, you will see a screen asking you to rate how you feel about this transfer. You will click a button, rating your feelings, from negative to positive. Then there will be a waiting period of about 10 seconds, before you see the next transfer.

24

Example 2: You might see this screen for 9 seconds:


VOLUNTARY.

Change to

your account:


account:

- $30

+ $30

A bit later, a message will be added to the lower screen half:


VOLUNTARY.

Change to

your account:


account:

- $30

+ $30

Approve

Reject

The buttons at the bottom will randomly switch locations. Once you have made a decision, it is final - you can’t go back. After you make a decision, you will see a screen asking you to rate how you feel about this transfer. Then there will be a waiting period of about 10 seconds, before you see the next transfer.

25

Quiz:

Before we begin, we want to ask a few questions to make sure you understand the experiment. We will ask you to determine your final payoff under different conditions. Remember, the experiment starts with you having $100 in your account, and $0 in the account for Food for Lane County (FFLC). So be sure you add these amounts to the changes below, to calculate the final amount. EXAMPLE: Suppose the following two transfers are randomly picked by the computer:

Mandatory: Change to your account: -$15 Change to FFLC account: +$15 Voluntary: Change to your account: -$30 Change to FFLC account: +$30

And suppose you had hit the approve button for the voluntary transfer.

Then I would get $100 - $15 - $30 = $55. FFLC would get $0 + $15+ $30 = $45. Now suppose you had hit the reject button for the voluntary transfer.

Then I would get $100 - $15 - $0 = $85. FFLC would get $0 + $15 + $0 = $15. QUIZ: 1. Suppose the following two transfers are randomly picked by the computer:



Then I would get ____________________ FFLC would get ____________________ Now suppose you had hit the reject button for the voluntary transfer.

Then I would get ____________________ FFLC would get ____________________

26

2. Suppose the following two transfers are randomly picked by the computer:

Mandatory: Change to your account: $0 Change to FFLC account: $0 Voluntary: Change to your account: -$30 Change to FFLC account: +$15



Then I would get ____________________ FFLC would get ____________________ 3. Suppose the following two transfers are randomly picked by the computer:

Mandatory: Change to your account: -$15 Change to FFLC account: +$30 Voluntary: Change to your account: +$45 Change to FFLC account: $0



Then I would get ____________________ FFLC would get ____________________ 4. Suppose the following two transfers are randomly picked by the computer:




Then I would get ____________________ FFLC would get ____________________

The RA checks the quiz and reviews any errors with the participant. Participants who do not

understand the payoff structure after this explanation are paid for their time and dismissed.

Successful participants are taken to the imaging lab.

27

Imaging Lab Instructions:

The RA now takes the participant to the imaging lab. The imaging lab staff explains the imaging procedure and runs though how to use the button boxes.

There is practice using the boxes, with randomly generated numbers.

We then start the experiment. This information is read just before the experiment:

“Remember – one voluntary and one mandatory transfer will be randomly chosen and played for

real money. Since you don’t know which one is chosen, your best strategy is to think carefully

about each transfer, and treat it as it is the one that counts. We are interested in recording your

brain response while you are considering these transfers, so please think carefully about the pros and cons of each transfer while you are considering it. Please do this even for the

mandatory transfers.

Remember – all your decisions are final. You cannot change your decisions once you have

clicked the buttons.”

supporting online material for - science...supporting online material for neural responses to...

Documents