Ol

AD

a

ARR2AA

KAEMPSS

1

Rebhrtr

hlW

Cf

0h

Behavioural Processes 104 (2014) 72–83

Contents lists available at ScienceDirect

Behavioural Processes

jo ur nal home p ag e: www.elsev ier .com/ locate /behavproc

n the dynamics of stimulus control during guided skillearning in nonhumans�

lliston K. Reid ∗, Nathan Folks, Jordan Hardyepartment of Psychology, Wofford College, Spartanburg, SC 29303, USA

r t i c l e i n f o

rticle history:eceived 21 August 2013eceived in revised form3 December 2013ccepted 6 January 2014vailable online 24 January 2014

eywords:utonomyxpertiseotor skill

a b s t r a c t

This study measured skill acquisition in the presence and absence of guiding cues in pigeons.It asked whether the speed of development of autonomy for the motor skill is influenced by the difficulty

level of two guiding-cue conditions requiring the same left–right response sequence. The Follow-Red con-dition required a simple go, no-go discrimination (red = S+, green = S−), whereas the Red–Green conditionwas a more difficult simultaneous chain requiring sensitivity to the serial order of key colors (red = S+,green = S− for the first peck, but red = S−, green = S+ for the second peck). Pigeons exposed to the difficultRed–Green condition displayed significantly higher accuracy levels during no-cues conditions earlier intraining than those exposed to the easier Follow-Red condition. A modified Power Law of Practice wasused to evaluate the null hypothesis that autonomy develops equally in explicit guiding-cues conditionsand no-cues conditions. This hypothesis was retained in the Follow-Red condition but rejected in the

rompt dependencekill learningtimulus control

Red–Green condition. Practice completing the response sequence in the Follow-Red and no-cues condi-tions both contributed equally to autonomy. Autonomy developed faster in the Red–Green group in bothconditions, and it developed unexpectedly rapidly during the second guiding-cues condition, implyingthe involvement of a second process for the Red–Green condition. We discuss the implications of theseresults to prompt dependence in children with learning disabilities, the transfer of stimulus control, andpotential behavioral interventions.

© 2014 Elsevier B.V. All rights reserved.

. Introduction

There are many ways of conceptualizing behavior patterns.esearch in this lab has focused on the rules of integration ofnvironmental cues and responses to produce adaptive patterns ofehavior. Behavior analysis has a long tradition of conceptualizingeterogeneous behavior patterns as behavior chains, in which eachesponse in the chain is presumed to be controlled by a discrimina-ive stimulus, and response-produced stimuli both reinforce thatesponse and “set the occasion” for the next response.

� We thank Sydney Kline, Carrie Martin, and N. Hunter Rackett for their reliableelp in conducting the experiment. We also thank Elizabeth Kyonka and Phil Hine-

ine for fruitful discussions and comments about this research. We are grateful toofford College for funding.∗ Corresponding author at: Department of Psychology, Wofford College, 429 Northhurch Street, Spartanburg, SC 29303, USA. Tel.: +1 864 597 4642;ax: +1 864 597 4649.

E-mail address: reidak@wofford.edu (A.K. Reid).

376-6357/$ – see front matter © 2014 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.beproc.2014.01.017

1.1. Stimulus control in motor skills

In a series of experiments, Reid et al. (2010) and Reid et al.(2013b) expanded this view of behavior chains by arguing thatbehavior chains in nonhumans are often equivalent to motor skillsin humans. Most people recognize the impressive skills demon-strated by dogs in agility training, dolphins and sea lions performingin amusement parks, and the many videos of trained animals avail-able on online sources. Animal training is an art of applying knownprinciples of learning and behavior. The typical focus is on the roleof reinforcement on behavior patterns, but we focus on the roleof stimulus control in the acquisition of a motor skill for two rea-sons. First, rats and pigeons are sometimes remarkably insensitiveto informative stimuli that should, on face value, come to controlresponding (Fox et al., 2014; Reid et al., 2013b). Second, childrenwith autism or severe learning disabilities often show “promptdependence”. They fail to learn to produce these skills indepen-dently, without continued prompts provided by the instructor.Prompt dependence describes the failure of stimulus control to

transfer from the teacher’s prompt (now say “Thank you”) to con-trol by situational cues (such as receiving a gift) (MacDuff et al.,2001). Foundational research about changes in stimulus controland cue interaction during skill learning should lead to deeper

dx.doi.org/10.1016/j.beproc.2014.01.017http://www.sciencedirect.com/science/journal/03766357http://www.elsevier.com/locate/behavprochttp://crossmark.crossref.org/dialog/?doi=10.1016/j.beproc.2014.01.017&domain=pdfmailto:reidak@wofford.edudx.doi.org/10.1016/j.beproc.2014.01.017

ral Pro

upi

1

atioSbfcitpchciptpse

1

dlagbtotlwaNrsstia‘ittwtthtfsa9mwst

Lights condition, control by practice cues was approximately equal

A.K. Reid et al. / Behaviou

nderstanding of the development of skill learning, the causes ofrompt dependence, and may help to suggest improved behavioral

nterventions.

.2. Two sources of stimulus control

Reid et al. (2010) emphasized how skill learning in rats requires change in stimulus control. At least two sources of stimulus con-rol are involved: environmental events from instructors or lightsn a Skinner box, and “practice cues” that result from the subject’swn behavior of repeating the same response pattern (Lattal, 1975;himp, 1981, 1982). As the skill is acquired, reliance on (or controly) practice cues increases until the behavior pattern can be per-ormed correctly and efficiently in the absence of explicit guidingues. High accuracy in the absence of the previous guiding cuess commonly called “autonomy” – the autonomous skill now con-rolled by newly-developed practice cues. The important role ofractice in skill learning has long been recognized in cognitive psy-hology even before Ebbinghaus (1885), but the variable of interestas often been the number of practice trials, rather than practiceues. Behavior analysis has much to offer to improve understand-ng of skill learning. It allows us to clearly identify and controlractice cues; it provides procedures for measuring the quantita-ive changes in control by environmental events and developingractice cues as motor skills are acquired; and it allows the mea-urement of cue interaction. These are the goals of the currentxperiment.

.3. Measuring practice cues and autonomy

Two conditions in Reid et al. (2010) provide a useful means ofefining the terms and skills used in the current study. Rats first

earned to “follow the light,” and then we removed the lights tossess how well they could complete the task without externaluidance (i.e., their degree of autonomy). The ‘skill’ to be acquiredy rats was a left–right (L–R) lever-press sequence in a discrete-rials procedure. The ‘guiding cues’ were the presence and absencef panel lights over the respective levers. At the beginning of eachrial in the ‘follow-the-light’ condition, the panel light over theeft lever was illuminated, whereas the light over the right lever

as off. A press to either lever caused the left light to turn offnd the right light to turn on. A second press terminated the trial.o feedback was provided within the trial about response accu-

acy. The L–R sequence produced a food pellet, whereas all otherequences produced timeout with the lights off. This responseequence was required in all conditions. Eleven rats were exposedo this condition until L–R sequence accuracy exceeded 80% with noncreasing or decreasing trends for five consecutive sessions. Once

rat achieved this accuracy-stability criterion, it was exposed to ano-lights’ condition in which both panel lights remained off, elim-nating the panel lights as cues to guide response selection withhe trial. The high 90% accuracy during the follow-the-lights condi-ion dropped to about 50% in the no-lights condition. Thus, ratsere able to complete the correct L–R sequence about half the

ime without the lights guiding response selection. We assumedhat their behavioral history of repeating the same L–R sequenceundreds of times allowed the development of “practice cues”hat were able to guide response selection at 50% accuracy. Dif-erent rats required different numbers of sessions to reach ourtability criteria. This allowed us to examine the size of the drops a function of the number of training sessions (ranging from

to 22 sessions). We observed an approximately linear relation:

ore training led to a greater ability to complete the sequenceithout the cues. We called this a “practice effect,” which demon-

trated the development of practice cues. More practice completinghe sequence led to greater autonomy, which we defined as the

cesses 104 (2014) 72–83 73

acquired ability (measured by accuracy level) to complete thesequence correctly without the lights as cues. Different amountsof training led to different accuracy levels, thus different degrees ofautonomy.

Subsequently, Reid et al. (2013b) repeated this basic procedurewith training on a more challenging “reversed-lights” guiding-cuecondition (Expt. 2). In this case, light off was S+, and light on wasS−, reversing the cues from Reid et al. (2010). The motor skill wasthe same L–R response sequence as before. This training conditionrequired about twice the number of sessions for rats to reach thesame accuracy-stability criteria as before, 28 sessions as comparedto 14. When switched to a no-cues condition that eliminated thelights as guiding cues, accuracy dropped only 20–25%. Consistentwith the idea of developing practice cues, more practice with theL–R sequence led to higher accuracy, i.e., more autonomy.

1.4. Cue interaction

Both of these studies provided important clues about the natureof cue interaction. Do explicit guiding cues and practice cues inter-act the same way as in Pavlovian conditioning? If behavior firstbecomes controlled by explicit cues (the panel lights as discrimi-native stimuli) and practice cues only develop later, should we notexpect to observe blocking of practice cues? Both studies demon-strated that practice cues developed while explicit guiding cueswere provided, and more exposure to the guiding cues led toimproved accuracy in the presence and in the absence of these cues.Reid et al. (2013b, Expt. 3) demonstrated that acquisition of theresponse sequence is delayed considerably when guiding cues arenot provided. These studies tentatively imply that explicit guidingcues facilitate the acquisition of practice cues, rather than competewith them for control of behavior.

Imagine the following scenario. You have moved to a new uni-versity, and you want your young child to learn how to walk fromthe parking lot to your new office. Which would promote fasterautonomy for your child: to “lead him or her by the hand” as youwalk along each sidewalk, or by providing less guidance such as ask-ing at each corner “which way do we go?” Reid et al. (2013a) askedwhether the development of control by practice cues is influencedby the degree of “effectiveness” of stimulus control by explicitguiding cues, with some cues being more effective at controllingbehavior than others. Accuracy may always be high if you “lead thechild by the hand” as the task is completed, but such direct guid-ance may not lead to faster autonomy. They also suggested thatprompt dependence may reflect this failure for the task to becomecontrolled adequately by other cues.

Following the demonstration by Reid et al. (2013b) that thelights and reversed-lights conditions differed reliably in their effec-tiveness as guiding cues, Reid et al. (2013a) asked whether controlby guiding cues and practice cues develop at the same rate, andwhether the effectiveness of guiding cues (Lights versus Reversed-Lights) influenced this rate. One group of rats acquired the L–Rsequence exposed to the Lights condition, and another group wasexposed to the Reversed-Lights condition. We measured develop-ing stimulus control by guiding cues and by practice cues inde-pendently in the same sessions by inserting probe trials withoutguiding cues. We found that while the Lights condition producedgreater accuracy when those cues were provided, the developmentof practice cues was retarded, with autonomy remaining low evenafter 36 sessions. Nevertheless, in the (less effective) Reversed-

to control by guiding cues across all 36 sessions. In this condition,control by both types of cues appeared to develop at approximatelythe same rate. In terms of the analogy, holding your child’s hand toomuch seems to slow the development of autonomy.

7 ral Pro

1g

tTespotciTaasrbtRluLpTsortgcrnsops

dRWin

2

2

oidbcc

2

ts2ra

4 A.K. Reid et al. / Behaviou

.5. Measuring skill acquisition in the presence and absence ofuiding cues in pigeons

The experiment presented here focused on these same ques-ions and was designed to extend the findings above to pigeons.he guiding cues in the experiments with rats had been the pres-nce and absence of illumination of panel lights. Hearst (1991)ummarized a large series of studies emphasizing the feature-ositive effect, which is the reduced stimulus control by the absencef a stimulus than by its presence. This could have been whyhe Reversed-Lights condition was less effective than the Lightsondition. By working with pigeons, we eliminated this potentialnfluence by asking pigeons to respond to red and green key lights.he purpose of this study was to measure the improvement ofutonomy (L–R accuracy during no-cues trials) with practice bylternating blocks of guiding-cues sessions with blocks of no-cuesessions. Our focus was on the nature of cue interaction as L–R accu-acy improves: Does autonomy develop differently within theselocks? As in Reid et al., 2013a, we exposed two groups of animalso conditions in which differed in acquisition of stimulus control.eid et al., 2013a manipulated the effectiveness of control by one

ight illuminated and another off over two levers. Here we manip-lated the complexity of the task by requiring pigeons to peck an–R sequence by “following” a key color that changed location, oreck the same L–R sequence in a simultaneous chain (Reid, 2009;errace, 1984, 2005) which required control by key color and byerial order. The easy Follow-Red condition with pigeons was anal-gous to the follow-the-lights condition for the rats. In this case,ed was S+ and green was S-. The first peck reversed the position ofhese colors on the two keys, without altering the response contin-ency. The more difficult Red–Green condition was a simultaneoushain requiring completion of a Red–Green sequence on the left andight keys, respectively. In this condition the colors of the keys didot change position as the L–R key-peck sequence was completed,o response selection had to become controlled by the serial orderf pecking: the red key was S+ and green key was S− for the firsteck, but they reversed roles (red key = S−, green key = S+) for theecond peck of the response sequence.

This experiment addressed two questions. Would autonomyevelop at different rates in the Follow-Red condition and theed–Green condition (i.e., would complexity of the task matter?)?ould exposure to the two types of guiding-cues conditions facil-

tate practice cues the same way, as measured in the next block ofo-cues sessions?

. Method

.1. Subjects

Eight experimentally naive homing pigeons (Columba livia),btained from Double T Farm (Glenwood, Iowa), served as exper-mental subjects. Supplementary mixed grain was provided afteraily sessions as necessary to maintain 85% of their free feedingody weights. The pigeons were housed individually in standardages with free access to water and grit under natural lightingonditions in an environmentally controlled animal facility.

.2. Apparatus

Four pigeon chambers were used, and each pigeon was placed inhe same chamber throughout the experiment. Two chambers were

tandard BRS-140 pigeon chambers (24 cm wide, 35.5 cm long, and9.5 cm high). The front panel was equipped with three 2.5-cmesponse keys (Med Associates ENV-123 with tri-color displays)rranged horizontally. Each key could be illuminated white, red, or

cesses 104 (2014) 72–83

green and was located 22 cm above the floor. A force of approxi-mately 0.15 N was necessary to operate each key. A food hopper(BRS/LVE grain magazine) centered below the center key 7.5 cmfrom the floor, provided access to mixed grain. The magazine wasilluminated when food was presented. Each experimental cham-ber was enclosed in a sound-attenuating box equipped with a 7-W120-V nightlight, located on the upper back wall of the box (behindthe operant chamber), and a fan that provided air circulation andmasked extraneous noise.

The other two chambers were Gerbrands-style pigeon chambers(35 cm wide, 43 cm long, and 36 cm high) were equipped with three2.5 cm response keys (Med Associates ENV-124 with multicoloreddisplays) in a row 23 cm above the floor. Each key required a forceof approximately 0.15 N to operate. A food magazine (Lehigh Val-ley Electronics) was centered below the center key 7.5 cm from thefloor, and the magazine was illuminated when food was presented.A 28-V house light (GE-1819) was located 6.5 cm above the cen-ter key. Ventilation fans masked extraneous noises. Each chambercontained a miniature color television camera located at the top ofthe back wall of the chamber to monitor subjects from an adjacentroom.

All operant chambers were controlled by a single Dell personalcomputer (Pentium 4) programmed in MED-PC IV, which controlledall experimental conditions and recorded every event and its timeof occurrence with 10-ms resolution.

2.3. Procedure

2.3.1. Phase 1An autoshaping procedure trained each pigeon to peck the cen-

ter key while illuminated white. Once pecking was reliable, thepigeons were randomly assigned to two groups of four pigeonseach, called the Follow-Red group and the Red–Green group. Thegroups differed in the manner in which key lights were used asguiding cues, as described below. The response requirements inthis discrete-trials procedure were identical for both groups and inall conditions. Food delivery always depended upon pecking onceon the left (L) key then once on the right (R) key. All other two-response sequences produced 6-s timeout in which all key lightswere extinguished. Except during autoshaping, the center key wasnot used. Completion of the L–R sequence activated the food hop-per for 3 s, followed by a 3-s intertrial interval (ITI). Key lights wereoff during food delivery and the ITI, but the houselight remained on.Sessions lasted a maximum of 45 min or until 60 reinforcers weredelivered. Subjects were exposed to this experimental procedureat approximately the same time each day, seven days per week.

The experiment was designed to alternate exposure to two con-ditions: one providing explicit guiding cues and one providing noguiding cues. In the no-cues conditions, key lights were either bothred or both green, which eliminated key color as a cue to guidethe L–R response sequence. Once accuracy was high (above 80%)and stable (no monotonic trend over the last five consecutive ses-sions) in the first exposure to a guiding cues condition, both groupswere further subdivided and exposed to either the Both-Red orBoth-Green conditions for five sessions. We defined a cycle as thecombination of five or more sessions of a guiding-cues condition(until accuracy exceeded 80% and appeared stable for five con-secutive sessions), followed by exactly five sessions of a no-cuescondition. All subjects were exposed to four complete cycles inorder to measure the development of L–R sequence accuracy inthe presence and absence of the guiding cues. Table 1 shows thenumber of sessions of exposure to each condition for each pigeon.

2.3.1.1. Follow-Red condition. In the Follow-Red condition, at thebeginning of each trial the left key was illuminated red and theright key illuminated green. Once a peck to either key occurred,

A.K. Reid et al. / Behavioural Processes 104 (2014) 72–83 75

Table 1Number of sessions each pigeon was exposed to each experimental condition.

Phase 1 Condition Pigeon

P220 P251 P274 P284 P308 P335 P355 P363

Cycle 1

Follow-Red 8 8 15 8 – – – –Red–Green – – – – 17 14 21 15Both-Green 5 5 – – 5 5 – –Both-Red – – 5 5 – – 5 5

Cycle 2

Follow-Red 5 6 12 7 – – – –Red–Green – – – – 11 9 10 6Both-Green 5 5 – – 5 5 – –Both-Red – – 5 5 – – 5 5

Cycle 3 Follow-Red 5 5 5 6 – – – –Red–Green – – – – 6 6 5 5Both-Green 5 5 – – 5 5 – –Both-Red – – 5 5 – – 5 5

Cycle 4

Follow-Red 5 5 5 7 – – – –Red–Green – – – – 5 10 6 5Both-Green 5 5 – – 5 5 – –Both-Red – – 5 5 – – 5 5

Follow-Red 5 5 5 5 – – – –Red–Green – – – – 6 5 5 5

– 5 5

bflTrofk

2blkanaippTbp

2tboe

2tr

2

wdsec

Phase 2 Follow-Red – – Red–Green 5 5 Both-White 5 5

oth key lights briefly pulsed off for 50 ms to indicate a success-ul key peck, and the colors on the two keys were reversed. Theeft key was illuminated green and the right key illuminated red.he second peck terminated the trial. Each L–R key peck sequenceesulted in food delivery, and all other sequences resulted in time-ut. The name of this condition indicates that pigeons could earnood by “following” the red light (S+) from the left key to the rightey without pecking keys illuminated green (S−).

.3.1.2. Red–Green condition. Each trial in the Red–Green conditionegan exactly as in the Follow-Red condition, with red and green

ights on the left and right keys, respectively. Once a peck to eitherey occurred, both key lights briefly pulsed off for 50 ms to indicate

successful key peck. However, the colors on the two keys didot reverse – they remained with red and green lights on the leftnd right keys, respectively. Each L–R key peck sequence resultedn food delivery, and all other sequences resulted in timeout. Thisrocedure, often called a simultaneous chain, required subjects toeck red then peck green on the left and right keys, respectively.hus the red key was S+ and green key was S− for the first peck,ut they reversed roles (red key = S−, green key = S+) for the secondeck of the response sequence.

.3.1.3. Both-Green condition. In the Both-Green condition, bothhe left and right keys were illuminated green. With the first peck,oth key lights pulsed off for 50 ms, returning to green. As in thether conditions, each L–R key peck sequence resulted in food deliv-ry, and all other sequences resulted in timeout.

.3.1.4. Both-Red condition. The Both-Red condition was identicalo the Both-Green condition except that both keys were illuminateded instead of green.

.3.2. Phase 2After completing the four cycles of Phase 1, we assessed how

ell each group would adapt to the alternative guiding-cues con-

ition, and then how well they would complete the L–R responseequence in the presence of a novel no-cues condition. We re-xposed pigeons in each group to their original guiding-cuesondition for five sessions to reestablish baseline accuracy. Then

– 5 5 5 55 – – – –5 5 5 5 5

we exposed each pigeon to the alternative guiding-cues conditionthey had not experienced before (i.e., subjects trained in the Follow-Red condition were exposed to the Red–Green condition and viceversa). Following five sessions of the unfamiliar guiding-cues con-dition, we exposed every subject to five sessions of a novel no-cuescondition in which both keys were illuminated white. This pro-cedure helped ascertain whether pigeons from different groupshad acquired different abilities to complete the response sequenceaccurately in the absence of explicit guiding cues.

2.3.2.1. Both-White condition. The Both-White condition was iden-tical to the Both-Red and Both-Green conditions except that thekeys were illuminated white. As in the other conditions, each L–Rkey peck sequence resulted in food delivery, and all other sequencesresulted in timeout.

3. Results

3.1. Initial acquisition

All pigeons in both groups acquired the L–R response sequence.As expected, acquisition in the Red–Green condition required moresessions than in the Follow-Red condition to reach the same accu-racy and stability criteria, t(6) = 2.99, p = 0.012 (one-tailed). As Fig. 1depicts, pigeons in the Follow-Red group reached our 80% accu-racy and 5-session stability criteria in 8–15 sessions (M = 9.75).Fig. 2 shows that the Red–Green group required 14–22 sessions(M = 17.5) to reach these criteria. This observation is consistentwith our prediction that acquisition of a simultaneous chain (theRed–Green condition) would be more challenging than learning asimpler Go, No-Go discrimination (red = S+, green = S−).

3.2. Accuracy across cycles

3.2.1. Follow-Red groupFor the Follow-Red group (see Fig. 1), each cycle consisted of a

Follow-Red condition followed by a no-cues condition, which wasBoth-Green for two birds and Both-Red for two birds. In the firstexposure to the no-cues condition in the first cycle, L–R key peckaccuracy dropped abruptly to approximately 25%, which would be

76 A.K. Reid et al. / Behavioural Processes 104 (2014) 72–83

L-R

Acc

urac

y (%

)

Sessions

0

20

40

60

80

100

0 5 10 15 20 25 30 35 40 45 50

Pigeon 220

Foll ow Re d BothGreen

Follow Red

BothGreen

Follow Red

BothGreen

Follow Red

BothGreen

Follow Red

0

20

40

60

80

100

0 5 10 15 20 25 30 35 40 45 50

Pigeon 25 1

Follow Red

Both Green

Follow Red

Both Green

Follow Red

Both Green

Follow Red

Both Green

Foll ow Red

0

20

40

60

80

100

0 5 10 15 20 25 30 35 40 45 50 55 60 65

Pigeon 27 4

BothRed

Both Red

Follow Red

Follow Red

Follow Red

Both Red

Follow Red

Both Red

Follow Red

0

20

40

60

80

100

0 5 10 15 20 25 30 35 40 45 50 55

Pigeon 28 4

FollowRed

FollowRed

BothRed

BothRed

Follow Red

Follow Red

BothRed

BothRed

Follow Red

F e Follon erent

e(FrsctPiHr(m

3

na2

ig. 1. This figure represents the L–R sequence accuracy for the four pigeons in tho-cues conditions. Pigeons represented in the two top panels were exposed to diff

xpected if the four possible sequences were selected randomlyas indicated by the horizontal dashed line). Upon returning to theollow-Red condition beginning the second cycle, accuracy quicklyecovered for all birds with P274 continuing to require more ses-ions to reach our accuracy criteria than the other birds. Acrossycles, sequence accuracy was reliably higher for all birds duringhe guiding cues condition than when those cues were eliminated.ercentage L–R accuracy in the no-cues conditions appeared toncrease across cycles for three birds (P274 being the exception).owever, this informal observation ignores variability in the accu-

acy levels observed in the prior Follow-Red guiding-cues conditionespecially for P274), so acquisition across cycles is discussed in

ore detail in Sections 3.4 and 3.5 below.

.2.2. Red–Green group

For the Red–Green group (see Fig. 2), in the first exposure to the

o-cues condition in the first cycle, L–R key peck accuracy droppedbruptly to approximately random chance for two birds, but to% and as much as 51% for the other birds. Upon returning to the

w-Red condition. Each panel shows the data from four cycles of guiding-cues and no-cues conditions than those in the bottom two panels.

Red–Green condition beginning the second cycle, accuracy quicklyrecovered for two birds, but gradually increased for P335 and P355over 4–5 sessions.

A central issue of this experiment, discussed in more detailbelow, concerns how performing the L–R key peck sequence in theguiding-cues conditions influences the pigeons’ ability to completethe same sequence in the absence of guiding cues. All four birdsshowed a substantial increase in accuracy from the last sessionof the first no-cues condition (cycle 1) to the first session of thesecond no-cues condition (cycle 2), separated by the interveningRed–Green condition. A repeated measures t-test comparing thesetwo points for the four pigeons demonstrated that this increase inaccuracy was statistically significant, t(3) = 11.44, p = 0.0014. Thisobservation demonstrates that the ability to complete the sequencecorrectly in the absence of guiding cues increased during the inter-

vening guiding-cues condition. This increase was not observedsystematically in the Follow-Red group (Fig. 1).

Across cycles, accuracy for P308 and P355 during the no-cues conditions was nearly as high as during guided-cue sessions,

A.K. Reid et al. / Behavioural Processes 104 (2014) 72–83 77

L-R

Acc

urac

y (%

)

Sessions

0

20

40

60

80

100

0 5 10 15 20 25 30 35 40 45 50 55 60 65

Pigeon 30 8

RedGreen

BothGreen

RedGreen

BothGree n

RedGreen

BothGreen

RedGreen

BothGreen

RedGreen

0

20

40

60

80

100

0 5 10 15 20 25 30 35 40 45 50 55 60 65

Pigeon 33 5

RedGreen

BothGreen

RedGreen

BothGreen

RedGreen

BothGreen

RedGreen

BothGreen

RedGreen

0

20

40

60

80

100

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70

Pigeon 35 5

BothRed

Both Red

RedGreen

RedGree n

RedGreen

Both Red

RedGree n

Both Red

RedGreen

0

20

40

60

80

100

0 5 10 15 20 25 30 35 40 45 50 55

Pigeon 36 3

RedGree n

BothRed

RedGreen

BothRed

RedGreen

BothRed

RedGreen

BothRed

RedGree n

F e Red–n erent

ahda

3

otitcsecTot

ig. 2. This figure represents the L–R sequence accuracy for the four pigeons in tho-cues conditions. Pigeons represented in the two top panels were exposed to diff

pproaching asymptote above 90%. Sequence accuracy was reliablyigher for only two birds (P335, P363) during the guiding-cues con-ition. Therefore, pigeons in this group varied considerably in theirbility to complete the sequence without guiding cues.

.3. Comparing Both-Red and Both-Green conditions

Did the two types of no-cues conditions have differential effectsn acquisition and the maintenance of sequence accuracy? Becausehese conditions may have multiple influences, we began by exam-ning the average size of the decrease in sequence accuracy fromhe guiding-cues condition to the no-cues condition across the fourycles for the Follow-Red group and the Red–Green group. Fig. 3hows the results of this analysis. The diagonal line would representqual-sized decreases in accuracy when switched from a guiding-

ues condition to a Both-Green condition or a Both-Red condition.hat is, we calculated the mean accuracy of the last five sessionsf the guiding-cues condition, and calculated the drop in accuracyo the first session (top panel) of the no-cues condition or drop to

Green condition. Each panel shows the data from four cycles of guiding-cues andno-cues conditions than those in the bottom two panels.

the average of five sessions (bottom panel) of the no-cues condi-tion. Points above the diagonal line would indicate that the drop inaccuracy was greater when switched to the Both-Green conditionthan during the switch to the Both-Red condition, and vice versafor points below the line. As one would expect during acquisition,drops generally decreased in size over the four successive cyclesof the procedure. As both panels show, the decreases in accuracyto the no-cues conditions did not deviate systematically from eachother. The decreases in accuracy when switched to Both-Red andBoth-Green were approximately equal.

3.4. Acquisition and cue interaction across cycles

Although Figs. 1 and 2 generally show increasing accuracyduring no-cues trials across the four cycles of the experiment,

variability (within- and between-subject) during the guiding-cues sessions complicates interpretation. Calculating L–R accuracyas a percentage of the previous averaged guiding-cue condition(as a proportion of baseline) helps clarify the acquisition curves

78 A.K. Reid et al. / Behavioural Pro

0

20

40

60

80

100

0 20 40 60 80 100

Drop

to B

oth

Gre

en (%

)To 1st No-Cues Session

Follow-Red

Red-Green

0

20

40

60

80

100

0 20 40 60 80 100

Drop

to B

oth

Gre

en (%

)

Drop to Both Red (%)

To Mean No-Cues Sessions

Fol low-Red

Red-Gree n

Fig. 3. This figure compares the two no-cues conditions to assess whether one con-dition produced larger drops in L–R accuracy than the other condition for each ofthe four cycles.The top panel compares the size of the drop in accuracy of the L–R sequence fromthe mean of the preceding guiding-cues condition to the first session of each no-cues condition. The bottom panel makes the same comparison to the mean of thefitc

awarocta

3p

rcFd2sspoAtm

ve no-cues sessions. Filled symbols represent the mean of the four pigeons inhe Follow-Red group, and open symbols represent the Red–Green group. In eachomparison, the largest drops occurred in the first cycle.

cross successive no-cues conditions. The central issue was howell the pigeons learned to perform the L–R key peck sequence

utonomously, when no explicit guiding cues were available to aidesponse selection. The top two panels of Fig. 4 show the percentagef mean baseline L–R accuracy for each of the successive no-cuesonditions. The data from the two pigeons in each no-cues condi-ion (Both-Green vs. Both-Red) were averaged together, since theyppeared equivalent.

.5. Modeling acquisition and cue interaction: modifying theower law of practice

We needed a principled way of estimating how much accu-acy should be expected to increase across successive no-cuesonditions. Recall that there was little if any difference in theollow-Red group between the last session of the no-cues con-ition in cycle 1 and the first session of that condition in cycle. However, the Red–Green group showed a large, statisticallyignificant difference. We wanted to know whether these acqui-ition patterns were consistent with a model that assumed thatractice improves accuracy independent of whether the practice

ccurred in the guided-cues condition or in the no-cues condition.lternatively, these two conditions might contribute differently

o autonomy in the no-cues condition. Nearly any simple modelight suffice as long as it does not distinguish between our two

cesses 104 (2014) 72–83

stimulus conditions, and it generates negatively accelerating curvesand (such as exponential or hyperbolic models). We exploreda model frequently used in cognitive psychology to account forthe improvement in skilled performance due to the accumulationof practice across trials. The Power Law of Practice (Newell andRosenbloom, 1981; Proctor and Vu, 2006; Rosenbloom, 2006) pro-poses that the reduction in reaction time (RT) that occurs withpractice during motor skill learning in humans can be describedby the equation

RT = A + BN−ˇ (1)in which A is the asymptotic RT after learning has been completed,B is performance time on the first trial, N is the number of practicetrials, and ̌ is the learning rate.

We measured increasing accuracy across sessions, notdecreasing reaction time across trials. Because the reciprocalof reaction time would represent a learning rate, we testedwhether the reciprocal of this equation would provide an adequatefit of our data, assuming that practice producing the L–R sequenceacross sessions would accumulate without regard to whetherthe practice occurred in the guiding-cues or no-cues conditions.Therefore, we calculated the best fitting parameter values in non-linear regression using the following equation to assess whetheracquisition curves would increase without regard to the actualstimulus condition:

Accuracy = 1A + BN−ˇ , (2)

By taking the reciprocal of Eq. (1) and applying Eq. (2) to ourdata, the meaning of each parameter may be altered. We appliedthe model to sessions containing many trials, rather than the actualnumber of practice trials (N). We set B to be the constant, 9.5 s/trial,which was the mean trial duration for all subjects in both groupsduring the first of the five guiding-cue sessions used as our stabilitymeasure in cycle 1.

Fig. 4 shows the best fit of this equation (using data onlyfrom the no-cues conditions) for the two no-cues conditions forthe Follow-Red group (top panel) and for the Red–Green group(middle panel). The model fit the learning curve nicely for the Both-Green birds in the Follow-Red group (R2 = 0.869) and nearly as wellfor the Both-Red birds (R2 = 0.624), generated by similar param-eter values (Both-Green: A = 0.015, ̌ = 3.269; Both-Red: A = 0.017,

̌ = 3.311). The best-fitting values for the Red–Green group reflectedmuch faster acquisition and earlier asymptote for both no-cuesconditions (Both-Green: A = 0.015, ̌ = 4.015; Both-Red: A = 0.012,

̌ = 3.649). The fit was better for the Both-Red birds (R2 = 0.690)than for the Both-Green birds (R2 = 0.380). The abrupt increases anddecreases across successive no-cues conditions in the Red–Greengroup were not observed in the Follow-Red group. Therefore, thissingle-learning process model more adequately accounts for acqui-sition during the Follow-Red condition than during the Red–Greencondition, which seemed to involve an additional process.

The bottom panel of Fig. 4 provides further insights into the dif-ferences between the Follow-Red and Red–Green groups. Pointsrepresent the averaged data from all four pigeons in each group.Eq. (2) accounts nicely (R2 = 0.832) for the acquisition curve in theFollow-Red group (filled symbols), with little deviation betweenobtained and expected accuracy values at the end and begin-ning of successive no-cues conditions. The fit of Eq. (2) to theRed–Green group depicts the rapid acquisition during the first cycleand the observed early asymptote. The fit of this equation is reason-able (R2 = 0.622), and the model does account for the observation

that accuracy was substantially higher in the second cycle for theRed–Green group than for the Follow-Red group, as verified bya mixed ANOVA, F(1,6) = 6.55, p = 0.043. However, the Red–Greengroup showed a sizeable unexplained deviation between obtained

A.K. Reid et al. / Behavioural Processes 104 (2014) 72–83 79

P

erce

ntag

e of

Mea

n B

asel

ine

L-R

Acc

urac

y0

25

50

75

100Follow-Red Pigeons

Both-Gre en Both-Red

Follo w Red

NoCues

Follow Red

Follow Red

Follow Red

NoCues

NoCues

NoCues

0

25

50

75

100Red-Green Pigeons

Red-Gree n

NoCues

NoCues

NoCues

NoCues

Red-Green

Red-Green

Red-Green

0

25

50

75

100

0 5 10 15 20 25 30 35 40

Mean Follow-Red Group Mean Red-Gree n Group

GC NoCuesNo

CuesNo

CuesNo

CuesGC GC GC

Sessions

F f the mt ns. Thr repre

aeifiFira

3

eghfsttr(

euget

ig. 4. This figure displays L–R accuracy in the no-cues conditions as a percentage ohe average accuracy for the two pigeons in the Both-Green or Both-Red conditioepresents the best fit of Equation 2, the modified Power Law of Practice. Error bars

nd expected accuracy values between the first and secondxposures to the no-cues conditions (as described in Section 3.2.2),ndicating that a second process may be involved. The values of bestt for the Red–Green group were: A = 0.013, ̌ = 3.795; and for theollow-Red group: A = 0.016, ̌ = 3.288. Even though the values of An every case were low with little variation across conditions, theeciprocal (1/A) ranged from 58.8 to 83.3, reflecting the differentsymptotes in accuracy observed by cycle 4 across conditions.

.6. Analysis of errors in no-cues conditions

Fig. 4 showed that accuracy during no-cues conditions was gen-rally higher for the Red–Green group than for the Follow-Redroup. Therefore, the overall frequency of incorrect sequences wasigher for the Follow-Red group. The top panel of Fig. 5 depicts the

requency of each type of sequence error made during the 20 ses-ions of exposure to the no-cues conditions for each group. Whereashe number of left–left (LL) errors was approximately the same forhe both groups, the Follow-Red group produced significantly moreight–left (RL) errors, t(158) = 5.43, p < 0.001, and more right–rightRR) errors than did the Red–Green group, t(158) = 3.38, p < 0.001.

Values in the top panel were naturally influenced by the differ-nce in total number of errors produced by each group. It is also

seful to compare which types of error were made by each group,iven that an error occurred, independent of the total number ofrrors made by each group. Therefore, the bottom panel depictshe proportion of each error type produced for each group (i.e.,

ean accuracy of the preceding guiding-cues condition. The top two panels displaye bottom panel displays the mean of all four pigeons in each group. Each curvesent SEMs.

number of each error type divided by the total number of errorsfor that group). The Follow-Red group was less likely to make LLerrors than was the Red–Green group, t(158) = 1.98, p = 0.0494.However, the Follow-Red group was more likely to make therelatively infrequent RL errors than was the Red–Green group,t(158) = 3.51, p < 0.001. The likelihood of RR errors was high forboth groups, but they did not differ reliably between the groups.

3.7. Phase 2: switch to the alternative guiding-cues condition andto a novel no-cues condition

3.7.1. Performance on the alternative guiding-cues conditionFollowing completion of all four cycles of Phase 1, we assessed

how well each group would adapt to the alternative guiding-cues condition, and then how well they would complete the L–Rresponse sequence in the presence of a novel no-cues condition.We re-exposed each subject to its original guiding-cues conditionfor five sessions to reestablish baseline accuracy. Each subject wasthen exposed to the alternative guiding-cues condition for fivesessions. Following five sessions of the unfamiliar guiding-cuescondition, we exposed every subject to five sessions of a novel no-cues condition in which both keys were illuminated white. Thisprocedure helped ascertain whether pigeons from different groups

had acquired different abilities to complete the response sequenceaccurately in the absence of explicit guiding cues.

Fig. 6 shows L–R accuracy under each of these conditions forindividual pigeons and group averages. All pigeons in each group

80 A.K. Reid et al. / Behavioural Pro

Fig. 5. The top panel compares the frequencies of each type of unreinforcedsequence (errors) for the Follow-Red and Red–Green groups that occurred dur-ing the 20 no-cues sessions. Three types of errors could occur in the two-responsesequences: left–left (LL), right–left (RL), and right–right (RR). The bottom panel elim-inates the influence of difference in total errors observed for each group by depictingthe proportion of each error type for each group, given that an error occurred. Aster-ib

rstoie(asangfwwntro

3

nhgiew

sks represent statistically significant differences between groups ( ̨ = 0.05). Errorars represent SEMs.

eestablished baseline accuracy and stability in the minimum fiveessions. For pigeons previously trained on the Follow-Red condi-ion (left panel), the switch to the Red–Green condition produced,n average, a substantial decrease in L–R accuracy, and accuracyncreased gradually and systematically over the five sessions. How-ver, for the pigeons previously trained on the Red–Green conditionright panel), the switch to the Follow-Red condition produced

small, transient decrease in accuracy. A mixed ANOVA demon-trated a significant main effect of group, F(1, 6) = 9.90, p = 0.020,nd a main effect of sessions, F(4, 24) = 12.96, p < 0.001. A sig-ificant group × sessions interaction demonstrated that the tworoups adapted to the alternative guiding-cues condition at dif-erent rates, F(4, 24) = 4.73, p = 0.006. This observation is consistentith the hypothesis that transition from Follow-Red to Red–Greenould be challenging because the simultaneous chain requiredovel control by the serial order in the response sequence, whereashe transitions from Red–Green to Follow-Red would involve aeduction or relaxation of previously acquired control by serialrder.

.7.2. Performance on the novel no-cues conditionWhen shifted to the novel Both-White condition that elimi-

ated the guiding cues, the Red–Green group produced slightlyigher average accuracy levels (M = 77.3%) than the Follow-Red

roup (M = 68.7%), but this difference was not statistically signif-cant, F(1,6) = 1.199, p = 0.316. After 53–72 sessions of previousxperience producing the L–R key peck sequence with andithout guiding cues, both groups were approximately equally

cesses 104 (2014) 72–83

capable of completing the sequence in the novel no-cues condi-tion.

4. Discussion

4.1. Comparing guiding-cue conditions

The main purpose of the current study was to determinewhether the speed of development of autonomy for a motor skill isinfluenced by the complexity of two guiding-cue conditions requir-ing the same L–R key peck sequence. The Follow-Red conditionrequired a simple go, no-go discrimination (red = S+, green = S−),whereas the Red–Green condition was a more difficult simulta-neous chain requiring sensitivity to the serial order of key colors(red = S+, green = S− for the first peck, but red = S−, green = S+ for thesecond peck). As the bottom panel of Fig. 4 shows, pigeons exposedto the difficult Red–Green condition displayed significantly higherlevels of autonomy during no-cues conditions earlier in trainingthan those exposed to the easier Follow-Red condition. Phase 2confirmed that the Red–Green condition was the more challengingbecause the Follow-Red group had more difficulty adjusting to thenovel Red–Green condition than vice versa. The speed of develop-ment of autonomy for a motor skill was influenced by the complexityof the two guiding-cue conditions.

This result confirms and extends the findings of Reid et al.(2013a) which showed that the speed of development of auton-omy for the L–R lever press sequence was influenced by theeffectiveness of stimulus control by explicit guiding cues duringacquisition. Using rats, they utilized two sets of guiding cues (Lightsvs. Reversed-Lights conditions) that had been previously demon-strated (Reid et al., 2013b) to differ in their effectiveness duringacquisition to control the L–R lever press sequence. These guiding-cue conditions utilized the presence or absence of illumination ofpanel lights to guide response selection. Accuracy in the presenceof the guiding cues was acquired faster in the Lights condition thanin the Reversed-Lights condition, just as we observed that acquisi-tion was faster during the first cycle in the Follow-Red than in theRed–Green conditions. Nevertheless, when the guiding cues wereeliminated in probe trials, subjects in the Reversed-Lights condi-tion were better able to complete the L–R sequence autonomouslythan those in the Lights condition. The less effective guiding cue(Reversed-Lights) produced greater levels of autonomy than didthe more effective cue (Lights), even though control by this guidingcue developed slower. The similarity in results between the twoexperiments demonstrates that the results with rats (Reid et al.,2013a) were not due to unique differences between the presenceand absence of lights, as feature-positive discrimination bias wouldsuggest (Hearst, 1991). The use of red and green key lights in thepresent experiment eliminated the possibility of feature-positivebias, yet resulted in the same observation that the more challengingguiding-cue condition produced greater autonomy.

4.2. Developing expertise

An intuitive application of the results from both experimentsis that guiding your child by the hand too much may reducehis or her ability to complete the task independently. Reid et al.(2013a) pointed out that the observation that a more challeng-ing guiding-cue condition produced greater levels of autonomythan did an “easier” condition is consistent with several expla-nations of human learning in the cognitive psychology literature.

However, differences in terminology and procedures make the rel-evance of this research to non-humans questionable. Craik andLockhart (1972) proposed a depth-of-processing explanation ofwhy elaborative rehearsal is superior to maintenance rehearsal.

A.K. Reid et al. / Behavioural Processes 104 (2014) 72–83 81

Fo llow-Red Re d-Green Both- Whit e Red-Gree n Fol low-Re d Bot h-Whit e Mea n

Trained on Follow-R ed Tra ined on Red-Green

L-R

Acc

urac

y (%

)

Sessions

Me an

Follow-Red Red-Green Both- White Red-Gree n Fol low-Red Bot h-Whit e

0

20

40

60

80

100 Pigeon 220

0

20

40

60

80

100 Pigeon 251

0

20

40

60

80

100 Pigeon 27 4

0

20

40

60

80

100 Pigeon 28 4

0

20

40

60

80

100 Pigeo n 308

0

20

40

60

80

100 Pigeon 335

0

20

40

60

80

100 Pigeo n 355

0

20

40

60

80

100 Pigeo n 363

0

20

40

60

80

100

0 5 10 150

20

40

60

80

100

0 5 10 15

F d grout

Seorcttirrattt

ig. 6. This figure displays the L–R accuracy levels for each pigeon in the Follow-Rehe experiment. Error bars represent SEMs.

ince then, many researchers have claimed that development ofxpertise in music requires not just repetitive practice, but elab-rative practice. This explanation would be consistent with ouresults if we assume that more attention occurred under the morehallenging Red–Green and Reversed-Lights conditions than underhe easier Follow-Red and Lights conditions. However, any rela-ionship between increased attention and “elaborative rehearsal”n rats and pigeons seems farfetched. We have no evidence thatats and pigeons engage in any type of rehearsal. Other animalesearchers are not so skeptical. Helton (2004, 2005, 2007a,b)

rgued that canines demonstrate “deliberate practice” in agilityraining. He argued that deliberate practice appears to be similaro elaborative rehearsal in Erricsson et al.’s (1993) theory of exper-ise development in humans which claims that deliberate practice,

p (left panels) and Red–Green group (right panels) and their averages in Phase 2 of

not genetics or talent, is the most important contributor to exper-tise acquisition. Perhaps future research will be able to discover therelation between a behavioral history producing stimulus controlby practice cues and the types of practice or rehearsal (deliberate,elaborative, maintenance) presumed to be involved in human skilllearning.

4.3. Modeling skill learning in nonhumans

What is the best way of modeling motor skill learning in non-

humans? One approach is to apply human models of expertise tononhumans and evaluate how well they fit. If the fit is good andthe model’s parameters remain meaningful, then nonhumans mayserve as animal models of human expertise. For example, Helton

8 ral Pro

(amhtrrtapittt

mPfgmtrAabctrgtbchoiiicmsumdRCetct

4

laaicimdhgoro

2 A.K. Reid et al. / Behaviou

2004, 2005, 2007a,b) concluded that canines given agility trainingppear to fit the early stages of Fitts and Posner’s (1967) three-stageodel of expertise development in humans. While this approach

as merit, it suffers by requiring the mapping of the cognitiveerminology described above to the behavior patterns of humbleats and pigeons. It is not clear that terms such as elaborativeehearsal, maintenance rehearsal, deliberate practice, and elabora-ive practice add anything to our understanding of rats and pigeonss they acquired the L–R motor skill in these procedures. A secondroblem is that the fit of human models of expertise might suffer

f humans have qualitatively different capabilities for skill learninghan those of rats and pigeons. For example, the human capacityo use tools and play musical instruments appears to differ qualita-ively, not only quantitatively, from that observed in other animals.

An alternative approach that we adopted here is to modify aathematical model of accumulated practice (the Power Law of

ractice) to evaluate how well it accounts for the observed dif-erences in skill accuracy between conditions providing explicituided cues and those that do not. By modifying the establishedodel, the parameters unfortunately lose their original interpre-

ation. However, the model assumed that practice completing theesponse sequence would accumulate equally in both conditions.s a model of cue interaction, it makes quantitative predictionsbout the amount of improvement to be expected during eachlock of each cycle. It served as the null hypothesis that practiceues develop equally in guiding-cues conditions and no-cues condi-ions. This hypothesis was retained in the Follow-Red condition butejected in the Red–Green condition. We observed that the fit wasood for the Follow-Red group, implying that practice completinghe response sequence in the Follow-Red and no-cues conditionsoth contributed equally to autonomy. Thus, the explicit guidingues did not block the development of practice cues. The modelelped demonstrate that autonomy in the Red–Green group devel-ped differently in two ways. Autonomy developed much fastern the Red–Green condition, and it produced an unexpected rapidncrease in accuracy during the second guiding-cues condition,mplying that a second process appeared to be involved for thishallenging Red–Green condition. We do not propose this simpleodel as a formal model of skill learning in nonhumans. Instead, it

erves as a useful assumption of how practice cues can accumulatentil a formal model is developed in future research. This futureodel must explain the paradoxical observation that practice cues

evelop more rapidly in more challenging conditions (the “difficult”ed–Green condition with pigeons in this study, and the Reversed-ue condition with less effective stimulus control with rats, Reidt al., 2013a). Eventually, this future model must also clearly iden-ify the set of stimuli in each experiment that we have simplyategorized as practice cues. Only then will we be able to answerhe question, “What is learned during motor skill learning?”

.4. Cue interaction

Nearly all models of Pavlovian conditioning predict that earlyearning of a predictive relationship will block or retard learning to

second stimulus appearing in combination, even though the firstnd second stimuli have the same temporal properties. Blockings usually explained by proposing competition between availableues. Motor skill learning occurs in procedures involving both thenstructor’s guiding cues and developing practice cues. Thus, one

ight expect early exposure to the guiding cues might block theevelopment of practice cues. The several experiments describedere, however, provide compelling evidence that early exposure to

uiding cues facilitates, not hinders, skill learning in the presencef those cues and the development of autonomy when the cues areemoved. A goal of future research should be to identify the naturef cue interaction during motor skill learning. We observe that

cesses 104 (2014) 72–83

practice cues can develop in the presence or in the absence ofexplicit guiding cues. We do not yet know what factors are responsi-ble for the learning rates in each condition. Is there an optimal timefor the instructor to leave and to ask the student to practice alone?What would we need to know in order to calculate this optimaldeparture time?

4.5. Prompt dependence

Prompt dependence in children with learning disabilities isobserved when stimulus control by the teacher’s prompt fails totransfer to contextual cues that should control that behavior. Typi-cally, the problem is not one of acquiring the motor skill; rather, it isa problem of the transfer of stimulus control. The procedures usedin Reid et al. (2010), Fox et al. (2014), and Reid et al. (2013b) differfrom the current experiment because they all trained a motor skilluntil it achieved accuracy and stability criteria before the transfer toa new stimulus condition. Some of these new conditions provideddifferent guiding cues, and some were no-cues conditions in whichthe skill should become controlled by newly developed practicecues in a similar context (within the same operant chamber, forexample). Procedures involving transfer to a new guiding-cues con-dition seem analogous to situations in which the prompt is nowprovided by a different person, whose prompt may differ slightly orsubstantially from the previous. Experiment 1 in Reid et al. (2013b)with rats and Fox et al. (2014) with pigeons demonstrated howtransfer from one guided-cues condition to another may trans-fer control rapidly, but simply reversing the order of exposure tothe same two conditions may prevent transfer of stimulus controlaltogether. Both studies also demonstrated that prior exposure toanother stimulus condition can alter, even reverse, these transfereffects. Procedures involving transfer to a no-cues condition maybe analogous to situations in which the prompt is eliminated andcontrol by endogenous or contextual cues is desired. Examples ofthis are the many situations in which the child should producethe behavior “on his or her own” in the proper context, such asdetecting when one should go to the toilet or how to calm down.

The methodology of these experiments offers much promiseto improve understanding of the causes of path dependence andhelping to suggest improved behavioral interventions to pro-mote improved transfer of stimulus control. The methodologyalso fits nicely with published research on prompt dependencewith handicapped children. For example, recognizing that severelyhandicapped children often become emotional in discriminationprocedures involving nonreinforcement or extinction, Touchette(1971) implemented a novel modification of Terrace’s (1963a,b)“errorless” discrimination procedure normally involving stimulusfading. He measured the transfer of stimulus control in hand-icapped children by imposing an accuracy-dependent temporaldelay at the beginning of each trial. During the delay, no guidingcues were provided, and he measured the child’s go/no–go dis-crimination. Each correct response increased the duration of thedelay, and each error decreased its duration. Touchette found thatdelaying the onset of the stimuli that initially controlled respon-ding, rather than gradually fading them out, brought respondingunder tight control of the temporal cues. All children learnedto respond before the explicit cue was provided. Touchette andHoward (1984) subsequently demonstrated that this procedure,called delayed prompting, is effective because anticipatory respon-ding (prior to the cue) provides increased relative reinforcementrate and produces rapid transfer of stimulus control. Since then,other researchers have used delayed-prompt procedures (Clark

and Green, 2004) or provided the most potent reinforcers forunprompted responses (Karsten and Carr, 2009). Procedures suchas delayed prompting could easily be implemented in the method-ology we have described for rats and pigeons.

ral Pro

cddeaaFbc

R

C

C

E

E

FF

HH

HH

HK

L

M

Touchette, P.E., 1971. Transfer of stimulus control: measuring the moment of trans-

A.K. Reid et al. / Behaviou

How do the current results inform instructional procedures forhildren with learning disabilities to prevent or eliminate promptependence? One implication of the current results is that promptependence might occur when the teacher’s guiding cues are soffective (“leading the child by the hand”) that too little attention isllocated to the desired endogenous or contextual cues that wouldllow autonomous performance of the skill in the new situation.uture research should explore whether prompt dependence cane reduced or prevented by practice with less effective or morehallenging guiding cues that require more focused attention.

eferences

lark, K.M., Green, G., 2004. Comparison of two procedures for teaching dictated-word/symbol relations to learners with autism. J. Appl. Behav. Anal. 37, 503–507.

raik, F.I.M., Lockhart, R.S., 1972. Levels of processing: a framework for memoryresearch. J. Verb. Learn. Verb. Behav. 11, 671–684.

bbinghaus, H., (H.A. Ruger, C.E. Bussenius, Trans.) 1885/1913. Memory: A Contribu-tion to Experimental Psychology. Columbia University, Teacher’s College, NewYork (Reprinted by Dover, New York, 1964).

rricsson, K.A., Krampe, R., Tesch-Romer, C., 1993. The role of deliberative practicein the acquisition of expert performance. Psychol. Rev. 100, 363–406.

itts, P.M., Posner, M.I., 1967. Human Performance. Brooks/Cole, Belmont, CA.ox A. E., Reid A. K., and Kyonka E. G. E. (in press). Behavioral history and pigeons’

“guiding cues” performance. The Psychological Record.earst, E., 1991. Psychology and nothing. Am. Sci. 79, 432–443.elton, W.S., 2004. The development of expertise: animal models? J. Gen. Psychol.

131, 86–96.elton, W.S., 2005. Animal expertise, conscious or not. Anim. Cognit. 8, 67–74.elton, W.S., 2007a. Deliberate practice in dogs: a canine model of expertise. J. Gen.

Psychol. 134, 247–257.elton, W.S., 2007b. Skill in expert dogs. J. Exp. Psychol. Appl. 13, 171–178.arsten, A.M., Carr, J.E., 2009. The effects of differential reinforcement of

unprompted responding on the skill acquisition of children with autism. J. Appl.Behav. Anal. 42, 327–334.

attal, K.A., 1975. Reinforcement contingencies as discriminative stimuli. J. Exp. Anal.

Behav. 23, 241–246.

acDuff, G.S., Krantz, P., McClannahan, L.E., 2001. Prompts and prompt-fadingstrategies for people with autism. In: Maurice, C., Burrows, K.E., Fritts, K.M. (Eds.),Making a Difference: Behavioral Intervention for Autism. PRO-ED, Austin, TX, pp.37–50.

cesses 104 (2014) 72–83 83

Newell, A., Rosenbloom, P.S., 1981. Mechanisms of skill acquisition and the law ofpractice. In: Anderson, J.R. (Ed.), Cognitive Skills and their Acquisition. Erlbaum,Hillsdale, NJ, pp. 1–55.

Proctor, R.W., Vu, K.-P.L., 2006. Laboratory studies of training, skill acquisition andretention of performance. In: Ericsson, K.A., Charness, N., Feltovich, P.J., Hoff-man, R.R. (Eds.), The Cambridge Handbook of Expertise and Expert Performance.Cambridge University Press, Cambridge, UK, pp. 265–286.

Reid, A.K., 2009. Resistance to change within heterogeneous response sequences.J. Exp. Psychol. Anim. Behav. Process. 35, 293–311, http://dx.doi.org/10.1037/a0013926.

Reid, A.K., DeMarco, G., Smith, K., Fort, T., Cousins, E., 2013a. The influences ofguiding cues on motor skill autonomy in rats. Learn. Behav. 41, 455–463,http://dx.doi.org/10.3758/s13420-013-0121-y.

Reid, A.K., Nill, C.A., Getz, B.R., 2010. Changes in stimulus control duringguided skill learning in rats. Behav. Process. 84, 511–515, http://dx.doi.org/10.1016/j.beproc.2010.01.001.

Reid, A.K., Rapport, H.F., Le, T.-A., 2013b. Why don’t guiding cues always guidein behavior chains? Learn. Behav. 41, 402–413, http://dx.doi.org/10.3758/s13420-013-0115-9.

Rosenbloom, P.S., 2006. A cognitive odyssey: from the power law of practice togeneral learning mechanism and beyond. Tutor. Quant. Methods Psychol. 2 (2),43–51.

Shimp, C.P., 1981. The local organization of behavior: discrimination of and memoryfor simple behavior patterns. J. Exp. Anal. Behav. 36, 303–315.

Shimp, C.P., 1982. On metaknowledge in the pigeon: an organism’s knowledge aboutits own behavior. Anim. Learn. Behav. 10, 358–364.

Terrace, H., 1963a. Discrimination learning with and without errors. J. Exp. Anal.Behav. 6, 1–27.

Terrace, H., 1963b. Errorless transfer of a discrimination across two continua. J. Exp.Anal. Behav. 6, 223–232.

Terrace, H.S., 1984. Simultaneous chaining: the problem it poses for traditionalchaining theory. In: Commons, M.L., Herrnstein, R.J., Wagner, A.R. (Eds.), Quan-titative Analyses of Behavior: Discrimination Processes. Ballinger, Cambridge,MA, pp. 115–138.

Terrace, H.S., 2005. The simultaneous chain: a new look at serially organizedbehavior. In: Wasserman, E.A., Zentall, T.R. (Eds.), Comparative Cognition: Exper-imental Explorations of Animal Intelligence. Oxford University Press, New York,pp. 481–514.

fer. J. Exp. Anal. Behav. 15, 347–354.Touchette, P.E., Howard, J.S., 1984. Errorless learning: reinforcement contingencies

and stimulus control transfer in delayed prompting. J. Appl. Behav. Anal. 17,175–188.

http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0005http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0010http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0015http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0020http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0025http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0025http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0025http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0025http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0025http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0035http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0035http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0035http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0035http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0035http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0035http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0035http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0035http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0035http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0040http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0045http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0050http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0055http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0060http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0070http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0065http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref1075dx.doi.org/10.1037/a0013926dx.doi.org/10.1037/a0013926dx.doi.org/10.3758/s13420-013-0121-ydx.doi.org/10.1016/j.beproc.2010.01.001dx.doi.org/10.1016/j.beproc.2010.01.001dx.doi.org/10.3758/s13420-013-0115-9dx.doi.org/10.3758/s13420-013-0115-9http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0100http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0105http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0110http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0115http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0120http://refhub.elsevier.com/S0376-6357(14)00031-X/sbref0125http://refhub.elsevier.