JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 2013, 9999, 1–15 NUMBER 9999 (XXX)

EFFECTS OF REINFORCER MAGNITUDE ON REINFORCED BEHAVIORAL VARIABILITY

ADAM H. DOUGHTY, KENNETH G. GIORNO, AND HANNAH L. MILLER
COLLEGE OF CHARLESTON

Eight pigeons were exposed to a two-component multiple schedule. In each component, four-peck
sequences across left and right keys were reinforced according to a variability threshold contingency. In one
condition, only infrequently occurring response sequences were reinforced in each component, thereby
generating highly variable sequences. In a separate condition, when the variability threshold contingency
was lenient in each component, sequences were much less variable. In each condition, reinforcer
magnitude was manipulated across components, and the larger reinforcer magnitude produced less
variability than the smaller reinforcer magnitude. These results suggest that larger reinforcers hinder the
reinforcement of behavioral variability. The results are interpretable in terms of the larger reinforcer
inducing a greater level of behavioral repetition, particularly as the time to reinforcement was approached.
This effect may have implications for reinforcing behavioral variability in humans.
Key words: behavioral variability, reinforcer magnitude, induction, pigeon, key peck

Bonem and Crossman (1988) comprehen- Variation in behavior usually is reduced as an

sively reviewed the literature involving the
effects of reinforcer magnitude on operant
behavior. They described many inconsistencies
in the literature and remarked, “…varying the
magnitude of a reinforcer is a very complex
operation” (p. 360). They also suggested that
this literature may benefit from more theoreti-
cally driven research. One approach to under-
standing the behavioral effects of reinforcer
magnitude is to posit that larger reinforcers may
not always strengthen or select behavior more
effectively than smaller reinforcers (e.g., Bizo,
Kettle, & Killeen, 2001; Doughty & Richards,
2002). The present research tested the view that
larger reinforcers sometimes exert their effects
by inducing greater behavioral repetition than
smaller reinforcers. Improving our understand-
ing of the effects of reinforcer magnitude on
behavioral variability is critical for several
reasons, including identifying better the benefi-
cial and detrimental effects of reinforcers on
problem solving and creativity (e.g., Ariely,
2010; Neuringer, 2003).
Portions of this research were supported by a SURF
(Summer Undergraduate Research with Faculty) grant as
well as a Faculty Research and Development grant, both
through the College of Charleston. The authors thank
several students for their contributions to this research,
particularly Nina Deese.
Address correspondence to Adam Doughty, Department
of Psychology, College of Charleston, 57 Coming Street,
Charleston, SC 29424 (doughtya@cofc.edu).
doi: 10.1002/jeab.50
organism approaches, in time or space, a
reinforcing stimulus (e.g., Akins, Domjan, &
Gutierrez, 1994; Gharib, Gade, & Roberts,
2004), even when behavioral variability is
reinforced (e.g., Cherot, Jones, & Neuringer,
1996). Cherot et al. exposed two groups of rats
to a fixed-ratio (FR) 4 contingency in which four
“successful” four-press sequences resulted in
reinforcement. For one (repeat) group, a four-
press sequence satisfied the FR requirement
only if it repeated one of the sequences from the
previous three trials, whereas for the other
(vary) group, a four-press sequence satisfied the
FR requirement only if it was different than each
of the last three sequences (i.e., Lag 3). Not
surprisingly, sequence variability was consider-
ably higher in the vary group. Importantly,
however, sequence variability decreased for
each group as the rats progressed through the
FR 4 contingency (e.g., sequence variability was
lowest in the fourth segment of the FR 4
contingency). This decreased variability yielded
more efficient responding (i.e., response se-
quences per reinforcer) for the repeat group
but lower efficiency for the vary group. One
interpretation of this finding is that behavioral
repetition was induced as the time to reinforce-
ment decreased, and the operant contingency
(i.e., vary or repeat) determined whether this
induced repetition was consistent with the
reinforced behavioral unit (cf. Baum, 2012).
As time to reinforcement approaches, larger
reinforcers may generate a greater level of
induced behavioral repetition than smaller

1
2 DOUGHTY et al.
reinforcers. In this context, one can compare
Reed and Wright (1988) with Doughty
and Richards (2002). Reed and Wright assessed
the effects of reinforcer magnitude (i.e., num-
ber of food pellets) on rats exposed to a chained
differential-reinforcement-of-low-rate (DRL)
variable-ratio (VR) schedule. Response rates
under the DRL and VR schedules decreased and
increased, respectively, with increases in rein-
forcer magnitude. The results are consistent
with the notion that a larger reinforcer better
strengthens or selects the reinforced behavioral
unit. Doughty and Richards examined the
effects of reinforcer magnitude on the behavior
of rats and pigeons under simple DRL schedules
(i.e., water amount to rats and duration of food
access to pigeons). The larger reinforcer
generated higher response rates, which reduced
considerably the number of obtained rein-
forcers (see also Kirshenbaum, Brown, Hughes,
& Doughty, 2008). These results are inconsistent
with the notion that a larger reinforcer always
strengthens or selects the reinforced behavioral
unit more effectively. A potentially important
difference across the studies is that the uncon-
ditional reinforcer immediately followed DRL-
schedule completion in Doughty and Richards,
while in Reed and Wright, DRL-schedule
completion immediately produced the stimulus
conditions correlated with the VR schedule.
Doughty and Richards (2002) interpreted
their findings within a framework that considers
the induced effects of reinforcers (e.g., Killeen,
1979, 1985; Segal, 1972; Staddon & Simmelhag,
1971). Under a simple DRL schedule, a
response immediately produces a reinforcer
only after a sufficiently long period of time
without the response. As such, this passage
of time can induce responding (e.g., Killeen,
1979; Lewis & Dougherty, 1992), with more
induced behavior in situations involving
larger reinforcers (e.g., Balsam, Brownstein, &
Shull, 1978; Doughty & Richards, 2002) or
greater reinforcer-deprivation levels (Lewis &
Dougherty, 1992). Greater induced behavioral
repetition under a DRL schedule can take the
form of a significantly larger number of
precriterion responses. Although Reed and
Wright interpreted their increased VR-schedule
response rates with larger reinforcers through
an enhanced response-strengthening process,
an account emphasizing greater induced repe-
tition by larger reinforcers is consistent with
their results.
It is easier to evaluate the position that larger
reinforcers induce greater behavioral repetition
by using a dependent measure other than
response rate. Stahlman and Blaisdell (2011)
assessed the effects of reinforcer magnitude on
behavioral spatiotemporal variation. Pigeons
were exposed to a discrete-trial procedure in
which a random-ratio (RR) 5 schedule was
presented after a 30-s intertrial interval (ITI).
There were five trial types, each signaled by a
distinct stimulus on a touchscreen. Four trial
types involved each combination of two levels of
reinforcer probability (12.5 and 4.4%) and
reinforcer magnitude (25.2 and 2.8 s of hopper
access). The final trial type involved a 100%
reinforcer probability and a 2.8-s reinforcer
magnitude, but the data from these trials were
omitted from analysis. Stahlman and Blaisdell
measured both spatial variation (i.e., where on
the touchscreen the pecks occurred) and
temporal variation (i.e., the time between
pecks). The larger reinforcer resulted in greater
behavioral repetition in both space and time,
and this effect was larger at the lower reinforcer
probability. The results support the conclusion
that larger reinforcers can induce behavioral
repetition to a greater extent than smaller
reinforcers. However, this behavioral repetition
was not inconsistent with occurrence of the
reinforced behavioral unit. A stricter test would
involve the reduction of behavioral variation by
larger reinforcers under a reinforcement con-
tingency requiring behavioral variability.
The purpose of the present experiment was to
determine if larger reinforcers reduce behav-
ioral variability under a reinforcement contin-
gency that requires behavioral variation.
Pigeons were exposed to a two-component
multiple schedule. In each component, four-
peck sequences across left and right keys were
reinforced according to a relative-frequency
threshold contingency (e.g., Denney &
Neuringer, 1998; Doughty & Lattal, 2001).
Under this contingency, a response sequence
results in a reinforcer only if its relative
frequency is less than a threshold value.
Threshold values were 0.05 and 0.30 in separate
conditions such that higher and lower levels of
behavioral variability were reinforced, respec-
tively. In each condition, reinforcer magnitude
differed across components (i.e., 2-s or 6-s access
to grain). It was predicted that behavioral
variability would be higher under the 0.05
contingency than the 0.30 contingency (e.g.,
 REINFORCER MAGNITUDE AND REINFORCED VARIABILITY
Grunow & Neuringer, 2002). Most importantly,
it was predicted that the larger reinforcer would
generate lower levels of behavioral variability
than the smaller reinforcer. Such a result would
extend the findings obtained by Cherot et al.
(1996) and Stahlman and Blaisdell (2011) and
support the claim that larger reinforcers can
induce greater behavioral repetition, even when
that repetition interferes with occurrence of the
reinforced behavioral unit.
Method
Subjects
Subjects were eight male pigeons obtained
from the Palmetto Pigeon Plant (Sumter, SC).
Five were White Carneau pigeons, approximate-
ly 3 yrs of age, with experimental histories of
reinforced behavioral variability (2196, 3776,
6067, 6117, and 9925). The other three pigeons
(0136, 2728, and 9222) were an “in-house
breed” (bred between White Carneau and Silver
King; labeled, Hubble), approximately 2 yrs of
age, with experimental histories unrelated to
reinforced behavioral variability. The pigeons
were maintained at 70% (0136, 2196, and 9222),
73% (2728 and 6067), or 80% (3776, 6117, and
9925) of their free-feeding body weights. The
weights lower than 80% were necessary to
maintain responding in prior experiments, so
there were no attempts to increase these weights
in this study. Each pigeon was housed individu-
ally, and water and health grit were available
continuously in each home cage. If necessary,
each pigeon was fed mixed grain at least 1 hr
after its session to maintain its target weight.
Lights were on in the room from 6:00 p.m. to
6:00 a.m., and sessions occurred Monday
through Friday.
Apparatus
Three identical three-key operant condition-
ing chambers were used. Each chamber had a
work area approximately 30 cm high by 25 cm
wide by 30 cm long and was enclosed in a sound-
attenuating box. White noise was presented
through a stereo near the chambers. Response
keys (2.5-cm diameter) were located on the
front wall and could be illuminated red or
green. Each of the two side keys was approxi-
mately 3 cm from the side wall and approxi-
mately 6 cm from the middle key. A white
houselight located in the upper portion of the
3
back panel provided general illumination. The
keylights and houselight were darkened during
reinforcement. Reinforcement was access to
mixed grain delivered in a food hopper located
behind a rectangular opening (approximately
6 cm high by 5 cm wide) centered on the bottom
of the front wall. A white light also was present
during food availability. Programming and data
recording were controlled by a computer in an
adjacent room using Med-PC IV software (Med-
Associates).
Procedure
Pretraining. Due to their more extensive
experimental histories, Pigeons 2196, 3776,
6067, 6117, and 9925 required no pretraining.
Pigeons 0136, 2728, and 9222 received one
session of magazine training, one session of key-
peck shaping, and approximately five sessions of
increasing FR-schedule requirements across the
two side keys. Specifically, one of the side keys
randomly appeared red or green and operated
until an FR schedule was completed. The FR-
schedule values increased until each of the four
options (i.e., left green, right green, left red,
right red) were completed under an FR 20
schedule.
Following pretraining, all pigeons were ex-
posed, for approximately 15 sessions, to proce-
dures that were nearly identical to the
procedures used in Condition 1 of the experi-
ment proper (see below). Each session began
with a 2-min period in which the key lights and
houselight were darkened. After this presession
blackout, one of the two multiple-schedule
components randomly started. The start of the
component was signaled by the onset of the
houselight and the two side keys. The side keys
were red in one component and green in the
other component. Regardless of key color, a
relative-frequency threshold contingency was in
effect such that a four-peck sequence was
eligible for food delivery only if its weighted
relative frequency was less than a threshold
value (see below). At the start of the component,
a pigeon could peck the left or right key. This
peck darkened both key lights for a 0.5-s
resetting delay; if a peck occurred to a dark
key, the 0.5-s delay restarted. After this delay,
both key lights turned on, and the pigeon could
peck either key again. This process continued
until four pecks had occurred. Immediately
after the fourth peck, the key lights and
4 DOUGHTY et al.
houselight were terminated and either food
delivery or a timeout occurred (i.e., the house-
light remained on across the four-peck
sequence).
Because there were two response options and
four required pecks, there were 16 possible
response sequences. The computer determined
after the fourth peck which of these sequences
occurred on a trial. The relative frequency of
that sequence was calculated by dividing the
number of times that sequence had occurred by
the total number of sequences emitted (sepa-
rate contingencies operated across components
such that the occurrence of sequences in one
component did not affect their calculation in
the other component). The relative frequencies
at the start of each session were taken from the
end of the preceding session. If the relative
frequency was greater than the threshold value,
a timeout occurred. If the relative frequency was
less than or equal to the threshold value, the
sequence was eligible for food delivery. During
this pretraining period, food only was delivered
following a “food-eligible” sequence if a variable-
interval (VI) 20-s schedule also had elapsed. If
the VI 20-s schedule had not elapsed, a timeout
occurred. Following food delivery, each of the
16 relative frequencies in the component was
multiplied by a weighting coefficient of 0.95 (see
Denney & Neuringer, 1998).
During this pretraining period, the threshold
value was 0.05 for four pigeons (2728, 3776,
6067, and 6117), and 0.30 for the other four
pigeons (0136, 2196, 9222, and 9925). The same
threshold value operated in both multiple-
schedule components. Across components,
different reinforcer magnitudes were arranged.
For Pigeons 0136, 2728, 6117, and 9925, the
duration of food delivery and timeout was 2 s in
the component with the red keys and 6 s in the
component with the green keys, whereas these
assignments were reversed for Pigeons 2196,
3776, 6067, and 9222. Component duration was
2 min with the restriction that a component
ended only after a food delivery or timeout. A
30-s intercomponent interval, in which all lights
remained off, separated multiple-schedule com-
ponents. The small and large reinforcer magni-
tude components strictly alternated until each
had occurred 10 times across the session.
Condition 1. The procedures in Condition 1
were identical to the aforementioned proce-
dures with the following three exceptions. First,
the VI 20-s schedule was eliminated. Every
response sequence that satisfied the threshold
contingency resulted in food delivery (see
Denney & Neuringer, 1998; Doughty &
Lattal, 2001). The change allowed the pigeons
to receive more frequent exposure to the
different reinforcer magnitudes. Second, ses-
sion duration was reduced to 10 multiple-
schedule components (compared to the previ-
ous 20 components). This change was made to
minimize potential satiation effects, given the
duration of food access in the component with
the larger reinforcer as well as the relatively high
obtained-reinforcement rates (see Table 1 be-
low). Third, the duration of food delivery in the
component with the smaller reinforcer was
increased from 2 to 3 s for Pigeons 2196 and
9222. This change was made to maintain
responding more effectively for these two
pigeons in this component.
Condition 1 continued for a minimum of 50
sessions and until responding stabilized across
nine consecutive sessions. These numbers were
selected because the session duration was
relatively brief (i.e., approximately 30 min).
Responding was judged to be stable, via visual
inspection, when there was minimal session-to-
session variability, and no trend, in either
multiple-schedule component for U-values
and response rates. U-value (Miller &
Frick, 1949; Page & Neuringer, 1985) is a
common measure of behavioral variability and
was calculated according to the formula: U ¼ - S
[(Rfi  log2 (Rfi)]/log2 (16)], where Rfi is the
unweighted relative frequency of each of the 16
response sequences. A U-value of 1 implies
complete uncertainty, or variability, and a U-
value of 0 implies complete certainty, or
repetition. Thus, higher U-values indicate
higher levels of between-sequence variability
(e.g., a U-value of 1 indicates that each of the 16
sequences occurred the same number of times).
Response rate was calculated by dividing the
number of sequences in a component by the
total amount of time spent in that component,
excluding time spent during reinforcer delivery
or timeout.
Condition 2. Condition 2 was identical to
Condition 1 with two exceptions. First, the
threshold value for each pigeon was reversed
(i.e., 0.05 to 0.30, or 0.30 to 0.05). Second, the
key colors correlated with the different rein-
forcer magnitudes were reversed (i.e., the
pigeons previously exposed to red and green
keys correlated with the small and large
 REINFORCER MAGNITUDE AND REINFORCED VARIABILITY 5

Table 1
For each pigeon in each condition, shown are the total number of sessions, response rate, reinforcement
rate, and proportion of reinforced sequences in each multiple-schedule component (i.e., Small and Large).
The means for these measures were calculated for the final stable sessions of each condition (with the
standard deviations in parentheses).

Responses/Min Reinforcers/Min Proportion of Reinforced Sequences

Condition Pigeon Sessions Small Large Small Large Small Large

0.05 0136 86 11.78 (0.65) 11.84 (0.52) 2.97 (0.79) 2.48 (0.63) 0.25 (0.07) 0.21 (0.06)
2196 78 8.73 (0.45) 9.27 (0.50) 2.66 (0.29) 2.82 (0.63) 0.31 (0.03) 0.31 (0.07)
2728 68 10.53 (0.39) 11.86 (0.57) 2.49 (0.39) 2.00 (0.43) 0.24 (0.04) 0.17 (0.04)
3776 59 10.66 (0.45) 9.92 (0.37) 2.92 (0.76) 2.41 (0.60) 0.27 (0.07) 0.25 (0.07)
6067 74 6.03 (1.16) 6.35 (0.71) 1.64 (0.61) 1.35 (0.33) 0.27 (0.09) 0.22 (0.06)
6117 127 7.90 (0.31) 6.82 (0.31) 2.38 (0.29) 2.14 (0.52) 0.30 (0.04) 0.31 (0.07)
9222 86 11.08 (0.48) 11.68 (0.55) 3.78 (0.31) 3.32 (0.75) 0.34 (0.04) 0.30 (0.07)
9925 87 13.22 (0.45) 13.63 (0.58) 3.26 (0.44) 3.39 (0.61) 0.25 (0.04) 0.25 (0.05)

0.30 0136 71 12.14 (0.40) 11.99 (0.50) 6.21 (1.26) 9.07 (0.78) 0.51 (0.07) 0.76 (0.06)
2196 103 7.22 (0.82) 7.19 (0.35) 4.81 (0.77) 5.43 (0.66) 0.67 (0.12) 0.76 (0.09)
2728 71 12.78 (0.70) 12.09 (0.81) 10.57 (1.12) 10.10 (1.17) 0.82 (0.11) 0.85 (0.05)
3776 134 10.24 (0.18) 10.69 (0.58) 8.32 (0.74) 9.38 (0.68) 0.81 (0.06) 0.88 (0.08)
6067 94 10.66 (0.59) 12.05 (0.59) 8.87 (0.57) 9.94 (0.93) 0.83 (0.03) 0.83 (0.09)
6117 60 8.77 (0.90) 8.76 (0.50) 7.85 (1.16) 7.53 (0.68) 0.90 (0.10) 0.87 (0.10)
9222 126 10.15 (0.41) 10.18 (0.91) 9.56 (0.64) 8.37 (0.86) 0.95 (0.07) 0.83 (0.11)
9925 98 11.37 (0.55) 12.17 (0.53) 9.10 (0.60) 8.80 (0.61) 0.80 (0.05) 0.72 (0.06)

reinforcers, respectively, were exposed to the
red and green keys correlated with the large and
small reinforcers and vice versa). Condition 2
continued for a minimum of 50 sessions and
until responding was judged to be stable over
nine sessions by visual inspection. This condi-
tion was terminated accidentally after eight
consecutive stable sessions for Pigeon 9222.
Results
Table 1 shows, for each pigeon, the number of
sessions in each condition as well as mean
response rate, mean reinforcement rate, and
mean proportion of reinforced sequences, and
their respective standard deviations, during the
final stable sessions of each condition in each
component. The proportion of reinforced
sequences was calculated by dividing the num-
ber of reinforced sequences in a component by
the total number of sequences that occurred in
that component. Visual inspection of response
rates of individual pigeons indicated that these
rates were not affected systematically by contin-
gency or reinforcer magnitude. A two-way
repeated-measures ANOVA (contingency, rein-
forcer magnitude) confirmed the absence of
significant differences (both main effects and
interaction p’s >.24). Visual inspection of rein-
forcement rates indicated that mean reinforce-
ment rate was higher with the smaller reinforcer
magnitude under the 0.05 contingency for six of
eight pigeons, whereas this result occurred
under the 0.30 contingency for only four of
the pigeons. However, neither the main effect of
magnitude nor the interaction was significant
(p’s >.19). The main effect of contingency was
significant, with reinforcement rates higher
under the 0.30 contingency, F(1, 7) ¼ 81.58,
p <.001. Visual and statistical inspections of the
proportion of reinforced sequences were similar
to the reinforcement-rate assessments. Neither
the main effect of magnitude nor the interac-
tion was significant (p’s >.27), but the main
effect of contingency was significant with a
higher proportion of reinforced sequences
under the 0.30 contingency, F(1, 7) ¼ 308.66,
p <.001.
Figure 1 shows the U-values in each compo-
nent in the final stable sessions under the 0.05
contingency for each pigeon. The left and right
columns of graphs display the U-values for the
pigeons that were exposed to the 0.05 contin-
gency in the first and second conditions,
respectively. Figure 2 shows the U-values in
each component in the final stable sessions
under the 0.30 contingency for each pigeon.
The left and right columns of graphs display the
U-values for the pigeons that were exposed to
the 0.30 contingency in the second and first
6 DOUGHTY et al.

Fig. 1. Displayed are U-values for each pigeon in each multiple-schedule component during the final stable sessions of the
0.05 contingency. The left and right columns display the U-values for pigeons that were exposed to the 0.05 contingency in
the first and second conditions, respectively. Reinforcer magnitude is differentiated by the circles and triangles.
 REINFORCER MAGNITUDE AND REINFORCED VARIABILITY 7

Fig. 2. Displayed are U-values for each pigeon in each multiple-schedule component during the final stable sessions of the
0.30 contingency. The left and right columns display the U-values for pigeons that were exposed to the 0.30 contingency in
the second and first conditions, respectively. Reinforcer magnitude is differentiated by the circles and triangles.
8 DOUGHTY et al.
conditions, respectively. The 0.05 contingency
generated higher U-values than the 0.30 contin-
gency, F(1, 7) ¼ 76.06, p <.001, and the smaller
reinforcer magnitude produced higher U-val-
ues than the larger reinforcer magnitude, F(1,
7) ¼ 19.09, p ¼.003. The interaction was not
significant (p ¼.20). In terms of observing the
differential effects of reinforcer magnitude in
individual pigeons, the difference in U-values
can be seen most clearly under the 0.05
contingency in Pigeons 2728, 0136, 6067,
9222, and 6117. Under the 0.30 contingency,
it can be seen most clearly in Pigeons 3776, 9222,
and 6117.
The results presented thus far indicate
that both the threshold contingency and
reinforcer magnitude impacted the degree of
sequence variation. To understand better how
these variables impacted behavioral variability,
Figures 3 to 6 display obtained relative-frequency
distributions as a function of reinforcer magni-
tude and threshold contingency. Specifically,
the relative frequency of each response se-
quence was calculated for the final stable
sessions of each condition in each component
(i.e., the number of times each sequence
occurred in a component divided by the total
number of sequences that occurred in the
component [without the weighting coeffi-
cient]). In each figure, the relative frequencies
are displayed according to a measure of within-
sequence variation. The leftmost eight sequen-
ces culminate with a changeover from one key to
the other (i.e., LRLR, RLRL, RLLR, LRRL,
RRLR, LLRL, RRRL, and LLLR). The next four
sequences culminate with two consecutive
responses to the same key (i.e., RRLL, LLRR,
LRLL, and RLRR). Finally, the next two
sequences culminate with three consecutive
responses to the same key (i.e., RLLL and
LRRR), and the last two sequences do not
contain a changeover (i.e., LLLL and RRRR).
In other words, the leftmost sequences culmi-
nate with within-sequence variation and the
rightmost sequences culminate with different
degrees of within-sequence repetition.
Figures 3 and 4 display these relative
response-sequence frequencies under the 0.30
contingency as a function of reinforcer magni-
tude for the pigeons that were exposed to this
contingency in the first and second conditions,
respectively. Across the small and large rein-
forcer magnitudes (columns), for all 16 cases, at
least three of the four most frequent sequences
are those sequences ending in three or four
consecutive pecks to the same key (i.e., RLLL,
LRRR, LLLL, and RRRR). Thus, the relatively
low levels of between-sequence variability under
the lenient 0.30 contingency resulted from the
dominance of sequences that culminated with
within-sequence repetition.
Figures 5 and 6 display the relative frequen-
cies under the 0.05 contingency as a function of
reinforcer magnitude for the pigeons that were
exposed to this contingency in the first and
second conditions, respectively. The distribu-
tions displayed in these two figures deviate
considerably from the distributions under the
0.30 contingency. As Figure 1 showed, the
effects of reinforcer magnitude on U-values
under the 0.05 contingency were observed most
clearly for Pigeons 2728, 0136, 6067, 9222, and
6117. Accordingly, the distributions for these
pigeons with the larger reinforcer were more
uneven than the distributions obtained with the
smaller reinforcer. The source of this uneven-
ness was that the larger reinforcer reduced the
likelihood of particular sequences to very low
levels. Specifically, across components for these
pigeons, there were a total of 15 sequences with
a relative frequency lower than .025. Fourteen of
these 15 sequences occurred with the larger
reinforcer, and 11 of those 14 sequences ended
in a changeover (i.e., ended in LR or RL). Thus,
although the 0.05 contingency maintained high
levels of between-sequence variation with each
reinforcer magnitude, the larger reinforcer
produced less behavioral variation by reducing
the likelihood of sequences that culminated in
within-sequence variation.
Discussion
The present experiment produced three
primary results. First, consistent with prior
findings (e.g., Grunow & Neuringer, 2002),
the 0.05 threshold contingency produced more
behavioral variability than the 0.30 threshold
contingency (e.g., higher U-values). Second,
differences in reinforcer magnitude also im-
pacted behavioral variability, with the larger
reinforcer producing less variability (e.g., lower
U-values). Third, the primary source of this
reduced behavioral variability by the larger
reinforcer was lower within-sequence variability
(indexed by changes in the relative-frequency
distributions). Thus, the present findings ex-
tend Cherot et al. (1996) and Stahlman and
 REINFORCER MAGNITUDE AND REINFORCED VARIABILITY 9

Fig. 3. Displayed are the relative response-sequence frequencies in each multiple-schedule component during the final
stable sessions of the 0.30 contingency for the pigeons that were exposed to the 0.30 contingency in the first condition.
10 DOUGHTY et al.

Fig. 4. Displayed are the relative response-sequence frequencies in each multiple-schedule component during the final
stable sessions of the 0.30 contingency for the pigeons that were exposed to the 0.30 contingency in the second condition.
 REINFORCER MAGNITUDE AND REINFORCED VARIABILITY 11

Fig. 5. Displayed are the relative response-sequence frequencies in each multiple-schedule component during the final
stable sessions of the 0.05 contingency for the pigeons that were exposed to the 0.05 contingency in the first condition.
12 DOUGHTY et al.

Fig. 6. Displayed are the relative response-sequence frequencies in each multiple-schedule component during the final
stable sessions of the 0.05 contingency for the pigeons that were exposed to the 0.05 contingency in the second condition.
 REINFORCER MAGNITUDE AND REINFORCED VARIABILITY
Blaisdell (2011) and show that larger reinforcers
can interfere with the reinforcement of behav-
ioral variability. This effect primarily was the
result of the larger reinforcer producing greater
within-sequence repetition, particularly at the
end of the sequence. The source of this result
can be interpreted as a function of the larger
reinforcer inducing greater behavioral repeti-
tion than the smaller reinforcer.
The present findings extend the literature
involving the effects of induced behavioral
repetition (e.g., Cherot et al., 1996; Stahlman
& Blaisdell, 2011). Stahlman and Blaisdell
reported greater behavioral spatiotemporal
repetition in pigeons with larger reinforcers,
however they did not differentially reinforce
behavioral variation. In the present experiment,
a larger reinforcer resulted in lower levels of
behavioral variability under a contingency
requiring behavioral variation. These detrimen-
tal effects of the larger reinforcer primarily
resulted from it reducing the probability of
sequences culminating in variation (i.e., se-
quences ending in LR or RL). Cherot et al.
observed induced behavioral repetition as time
to reinforcement was approached under a
variability reinforcement contingency. The
present results extend those findings by demon-
strating that this induced behavioral repetition
is greater with larger reinforcers. Further, the
results reported by Reed and Wright (1988) and
Doughty and Richards (2002) can be reconciled
if their respective VR-schedule and DRL-sched-
ule response-rate changes were due, at least in
part, to greater behavioral repetition induced by
larger reinforcers.
The operant contingency is one factor that
can determine whether greater induced behav-
ioral repetition by larger reinforcers is observed.
For example, Doughty, Galuska, Dawson, and
Brierley (2012) obtained enhanced lever-press
acquisition in rats under unsignaled delayed
reinforcement (i.e., a tandem FR 1 differential-
reinforcement-of-other-behavior [DRO] 30-s
schedule) with a larger reinforcer (i.e., six,
rather than one, food pellets). Induced lever
pressing by the larger reinforcer near the end of
the DRO requirement was eliminated, or at least
reduced considerably, by the relatively long
response–reinforcer delay created by the tan-
dem FR 1 DRO 30-s schedule (i.e., compare the
results to Doughty & Richards, 2002). As such,
this relatively strong operant contingency al-
lowed the larger reinforcer to exert its greater
13
response-strengthening, or selection, effects.
Future research can assess whether larger
reinforcers always produce lower levels of
reinforced behavioral variability, or whether
particular contingencies can minimize their
detrimental impact.
It is important to underscore the different
variability levels engendered by the threshold
contingencies in the present experiment. The
higher U-values observed under the 0.05
contingency replicates past research (e.g.,
Grunow & Neuringer, 2002) and demonstrates
the powerful role of reinforcement contingen-
cies in generating behavioral variation (e.g.,
Neuringer, 2002; Neuringer & Jensen, 2013).
Stated differently, although the larger reinforc-
er resulted in lower levels of behavioral variabil-
ity, relatively high levels of behavioral variability
were established under the 0.05 contingency
with each reinforcer magnitude. These latter
results are integral to consider in discussions
about the potential negative effects of promot-
ing creativity and problem solving in people
using reinforcement-based approaches (e.g.,
Ariely, 2010; Neuringer, 2003). That is, even if
the reinforcer-magnitude effects observed in
the present study can be extended to human
behavior, the necessity of using reinforcement
to maintain high overall variability levels cannot
be ignored.
Procedural limitations in the present work
should be noted. First, the absolute value of the
larger reinforcer was not studied beyond 6-s
access to grain so as to minimize potential
satiation effects. Larger reinforcers may have
exerted even greater repetition-inducing ef-
fects. Second, an enhanced impact of reinforcer
magnitude may be observed with a greater
difference between the two magnitudes (i.e.,
this difference was 4 s for six pigeons, and only
3 s for the other two pigeons). Third, the effects
of the larger reinforcer may have been reduced,
to some degree, by the manipulation of timeout
duration concurrent with reinforcer duration
(i.e., timeout duration in each component
matched the duration of food access in that
component). We manipulated timeout duration
concurrent with reinforcer duration to assess
the negative effects of larger reinforcers under
stricter conditions (i.e., conditions less likely
to demonstrate an effect). Neuringer (1991)
reported higher levels of response variation after
a longer time since the preceding response and
following timeout rather than reinforcement.
14 DOUGHTY et al.
Based on these results, the longer timeout
duration in the present experiment should
have increased behavioral variation more than
the briefer timeout duration, thus reducing the
repetition induced by the larger reinforcer.
Finally, our statistical analyses probably were
limited by the relatively small number of pigeons
used in this research.
There were no consistent response-rate differ-
ences in the present experiment as a function of
threshold value or reinforcer magnitude. The
absence of response-rate differences as a
function of threshold value may seem surprising
given the considerable discrepancy in reinforce-
ment rate across conditions. Grunow and
Neuringer (2002; Experiment 1) measured
both U-value and response rate as a function
of threshold value (i.e., 0.037, 0.055, 0.074, and
0.37, across groups) and reinforcement sched-
ule (i.e., FR 1 [as studied in the present
experiment], VI 1 min, and VI 5 min, across
conditions). They also observed minimal re-
sponse-rate differences in their FR 1 condition,
despite considerable U-value differences across
groups (e.g., the mean response-rate difference
across the 0.055 and 0.37 groups was two
sequences per min). The absence of a differen-
tial effect of reinforcer magnitude on response
rate in the present experiment also is not
without precedent (e.g., Bonem & Crossman,
1988). At this point, it only can be concluded
that reinforcer magnitude is more likely to affect
response-sequence variation than response-se-
quence rate, at least under the conditions
investigated here.
The present findings may be useful to consider
in the context of both reinforcement theory and
application. Behavior-analytic theorists increas-
ingly have questioned the typical response-
strengthening account of reinforcement (e.g.,
Baum, 2012; Shahan, 2010), including by placing
a greater emphasis on the induced effects of
reinforcers (e.g., Baum, 2012). A conceptualiza-
tion of the reinforcement process that includes
induction effects may benefit from considering
the present results as well as the general relation
between behavioral variation and schedule-
induced effects (e.g., Cherot et al., 1996; Lee,
Sturmey, & Fields, 2007; Neuringer & Jensen,
2013). Practically speaking, the reinforcement of
behavioral variability in people with and without
autism has received increased interest in applied
behavior analysis recently (e.g., Cammilleri &
Hanley, 2005; Harding, Wacker, Berg, Rick, &
Lee, 2004; Hopkinson & Neuringer, 2003;
Lee, McComas, & Jawor, 2002; Miller &
Neuringer, 2000; Napolitano, Smith, Zarcone,
Goodkin, & McAdam, 2010). Translation is
needed to understand fully how the present
results bear on that work; however, our findings
tentatively suggest that it may be easier
to reinforce behavioral variability in people
using smaller rather than larger reinforcer
magnitudes.
References
Akins, C. K., Domjan, M., & Gutierrez, G. (1994). Topogra-
phy of sexually conditioned behavior in male Japanese
Quail (Coturnix japonica) depends on the CS-US
interval. Journal of Experimental Psychology: Animal Behav-
ior Processes, 20, 199–209.
Ariely, D. (2010). The upside of irrationality: The unexpected
benefits of defying logic at work and at home. New York, NY:
HarperCollins Publishers.
Balsam, P. D., Brownstein, A. J., & Shull, R. L. (1978). Effect
of varying the duration of grain presentation on
automaintenance. Journal of the Experimental Analysis of
Behavior, 29, 27–36.
Baum, W. M. (2012). Rethinking reinforcement: Allocation,
induction, and contingency. Journal of the Experimental
Analysis of Behavior, 97, 101–124.
Bizo, L. A., Kettle, L. C., & Killeen, P. R. (2001). Rats don’t
always respond faster for more food: The paradoxical
incentive effect. Animal Learning and Behavior, 29, 66–78.
Bonem, M., & Crossman, E. K. (1988). Elucidating the effects
of reinforcement magnitude. Psychological Bulletin, 104,
348–362.
Cammilleri, A. P., & Hanley, G. P. (2005). Use of a lag
differential reinforcement contingency to increase
varied selections of classroom activities. Journal of Applied
Behavior Analysis, 38, 111–115.
Cherot, C., Jones, A., & Neuringer, A. (1996). Reinforced
variability decreases with approach to reinforcers.
Journal of Experimental Psychology: Animal Behavior Process-
es, 22, 497–508.
Denney, J., & Neuringer, A. (1998). Behavioral variability is
controlled by discriminative stimuli. Animal Learning &
Behavior, 26, 154–162.
Doughty, A. H., Galuska, C. M., Dawson, A. E., & Brierley,
K. P. (2012). Effects of reinforcer magnitude on
response acquisition with unsignaled delayed reinforce-
ment. Behavioural Processes, 90, 287–290.
Doughty, A. H., & Lattal, K. A. (2001). Resistance to change
of operant variation and repetition. Journal of the
Experimental Analysis of Behavior, 76, 195–215.
Doughty, A. H., & Richards, J. B. (2002). Effects of reinforcer
magnitude on responding under differential-reinforce-
ment-of-low-rate schedules of rats and pigeons. Journal of
the Experimental Analysis of Behavior, 78, 17–30.
Gharib, A., Gade, C., & Roberts, S. (2004). Control of
variation by reward probability. Journal of Experimental
Psychology: Animal Behavior Processes, 30, 271–282.
Grunow, A., & Neuringer, A. (2002). Learning to vary and
varying to learn. Psychonomic Bulletin & Review, 9, 250–258.
Harding, J. W., Wacker, D. P., Berg, W. K., Rick, G., & Lee,
J. F. (2004). Promoting response variability and stimulus
 REINFORCER MAGNITUDE AND REINFORCED VARIABILITY
generalization in martial arts training. Journal of Applied
Behavior Analysis, 37, 185–196.
Hopkinson, J., & Neuringer, A. (2003). Modifying behavioral
variability in moderately depressed students. Behavior
Modification, 27, 251–264.
Killeen, P. R. (1979). Arousal: Its genesis, modulation, and
extinction. In M. D. Zeiler, & P. Harzem (Eds.), Advances
in analysis of behaviour, Vol. 1 Reinforcement and the
organization of behaviour (pp. 31–78). Chichester,
England: Wiley.
Killeen, P. R. (1985). Incentive theory: IV. Magnitude of
reward. Journal of the Experimental Analysis of Behavior, 43,
407–417.
Kirshenbaum, A. P., Brown, S. J., Hughes, D. M., & Doughty,
A. H. (2008). Differential reinforcement-of-low-rate-
schedule performance and nicotine administration: A
systematic investigation of dose, dose-regimen, and
schedule requirement. Behavioural Pharmacology, 19,
683–697.
Lee, R., McComas, J. J., & Jawor, J. (2002). The effects of
differential and lag reinforcement schedules on varied
verbal responding by individuals with autism. Journal of
Applied Behavior Analysis, 35, 391–402.
Lee, R., Sturmey, P., & Fields, L. (2007). Schedule-induced
and operant mechanisms that influence response
variability: A review and implications for future inves-
tigations. The Psychological Record, 57, 429–455.
Lewis, P., & Dougherty, D. M. (1992). Pigeon performance
on a variable-interval omission schedule at different
levels of food deprivation. Behavioural Processes, 27,
27–36.
Miller, G. A., & Frick, F. C. (1949). Statistical behavioristics
and sequences of responses. Psychological Review, 56,
311–324.
Miller, N., & Neuringer, A. (2000). Reinforcing variability in
adolescents with autism. Journal of Applied Behavior
Analysis, 33, 151–165.
Napolitano, D. A., Smith, T., Zarcone, J. R., Goodkin, K., &
McAdam, D. B. (2010). Increasing response diversity in
children with autism. Journal of Applied Behavior Analysis,
43, 265–271.
15
Neuringer, A. (1991). Operant variability and repetition as
functions of interresponse time. Journal of Experimental
Psychology: Animal Behavior Processes, 17, 3–12.
Neuringer, A. (2002). Operant variability: Evidence,
functions, and theory. Psychonomic Bulletin and Review,
9, 672–705.
Neuringer, A. (2003). Reinforced variability and creativity.
In K. A. Lattal, & P. N. Chase (Eds.), Behavior Theory and
Philosophy (pp. 323–338). New York, NY: Kluwer
Academic/Plenum.
Neuringer, A., & Jensen, G. (2013). Operant variability.
In G. J. Madden, (Editor-in-Chief) W. V. Dube,
T. D. Hackenberg, G. P. Hanley, & K. A. Lattal (Associate
Editors). APA handbook of behavior analysis: Volume 1
(pp. 513–546). Washington, DC: American Psychologi-
cal Association.
Page, S., & Neuringer, A. (1985). Variability is an operant.
Journal of Experimental Psychology: Animal Behavior Process-
es, 11, 429–452.
Reed, P., & Wright, J. E. (1988). Effects of food reinforce-
ment on free-operant response rates. Journal of the
Experimental Analysis of Behavior, 49, 75–85.
Segal, E. F. (1972). Induction and the provenance of
operants. In R. M. Gilbert, & J. R. Millenson (Eds.),
Reinforcement: Behavioral analyses (pp. 1–34). New York:
Academic Press.
Shahan, T. A. (2010). Conditioned reinforcement and
response strength. Journal of the Experimental Analysis of
Behavior, 93, 269–289.
Staddon, J. E. R., & Simmelhag, V. L. (1971). The
“superstition” experiment: A reexamination of its
implications for the principles of adaptive behavior.
Psychological Review, 78, 3–43.
Stahlman, W. D., & Blaisdell, A. P. (2011). The modulation
of operant variation by the probability, magnitude, and
delay of reinforcement. Learning and Motivation, 42,
221–236.
Received: November 20, 2012
Final Acceptance: August 28, 2013