The free food (contrafreeloading) phenomenon: A reviewand analysis - ZOOSnippets
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Animal Leaming & Behavior
1977, 5 (3), 221-235
The free food (contrafreeloading) phenomenon:
A review and analysis
STEVE R. OSBORNE
Arizona State University, Tempe, Arizona 85281
Anima1s will perform an operant response to obtain food when abundant free food is availab1e.
These data have implications for current 1earning theories, especial1y in terms of the motiva-
tiona1 variables associated with such behavior. The present paper reviews the literature and
provides an analysis that suggests that responding for food in the presence of free food is im-
portantly controlled by stimulus change attendant upon response-dependent food presenta-
tion. This apparent stimulus-reinforcer effect on behavior is compared to that observed in other
areas of anima1 1earning research that include preference between schedu1es of response-
dependent and response-independent reinforcement, preference between schedu1es of signa1ed
and unsigna1ed reinforcement, autoshaping and automaintenance, and self-reinforcement in
anima1s.
When an animal is provided a choice between demonstrated with both water (Knutson & Carlson,
making an operant response to obtain food or by 1973; PoweIl, 1974; Robertson & Anderson, 1975;
eating it from a source of continuously available Taylor, 1972, 1975) and sucrose solution reinforcers
free food, a curious thing happens. Animals not only (Carder, 1972; Tarte, Townsend, Vernon, & Rovner,
acquire and continue to make the operant response, 1974). Although not food deprived, rats (Davidson,
but they often appear to "prefer" the response- 1971; Kopp, Bourland, Tarte, & Vernon, 1976;
dependent food. That animals respond for food in McLaughlin, Kleinman, & Vaughn, 1973; Tarte &
the presence of identical free food is by now a well- Synder, 1972), pigeons, (Bilbery et al. , 1973;
established finding in the experimentalliterature. Neuringer, 1969; Wallace, Osborne, Norborg, &
Behavior has been maintained on both fixed-ratio Fantino, 1973) and chickens (Duncan & Hughes,
(Alferink, Crossman, & Cheney, 1973; Atnip & 1972) have continued to respond for food in the
Hothersall, 1973; Carder & Berkowitz, 1970; presence of free food.
Davidson, 1971; Hothersall, Huey, & Thatcher, Two separate, but related, issues are involved in
1973; Tarte & Vernon, 1974) and variable-interval the free food phenomenon (0' Amato, 1974; Taylor,
(Bilbrey, Patterson, & Winokur, 1973; Neuringer, 1972). First, under what conditions, i.e., to what
1970; Rachlin & Baum, 1972; Sawisch & Denny, extent, will animals "work" für food when free food
1973) schedules of reinforcement when free food was is available? Second, do animals prefer to work for
concurrently available. Different strains of rats food when free food is available? In attempting to
(Atnip & Hothersall, 1973; Hothersall et al. , 1973; address the second issue, free food experiments fre-
Powell, i 974), rats with septal and ventromedial quently have been oriented towards specifying the
hypothalamic lesions (Singh, 1972a), mice (Pallaud, preference relationship between free and response-
1971), chickens (Duncan & Hughes, 1972), pigeons dependent food. Preference for a food source exists
(Neuringer, 1969, 1970), crows (PoweIl, 1974), cats when the proportion of food obtained from that
(Koffer & Coulson, 1971), gerbils (Lambe & Guy, source exceeds 50010 of the total food obtained during
1973), Siamese fighting fish (Baenninger, & a session. The data from studies of preference often
Mattleman, 1973), and humans (Singh, 1970; Singh vary enormously; some studies report a large pre-
& Query, 1971; Tarte, 1972) have been shown to ference for response-dependent food, others report
work for reinforcers when equivalent free reinforcers an equally large preference for free food (cf. Carder
were available. Although food reinforcers typically & Berkowitz, 1970; Taylor, 1972). Inconsistency of
have been used, the phenomenon also has been results from similar studies, coupled with difficulty
in specifying the variables that control free food re-
sponding, may be responsible for the paucity of
I thank L. Meyerson, S. Parkinson, and members of the "Glass attempts to integrate free food data into existing
Bead Gang,". especially P. Killeen, for their helpful comments
on an earlier draft of this paper, and Ada Cardenas for her help in
theory, or to relate it to other areas of research in
preparing tlre" manuscript, Reprints may be obtained from S. animal learning. However, recent research suggests
Osborne, Department of Psychology, Arizona Stare University, new perspectives that a110w an explanation of the
Tempe, Arizona 85281. phenomenon that not only is parsimonious with
221222 OSBORNE
regard to existing free food data, but one that is dency, Brown and Jenkins (1968) found that experi-
relevant for other research areas as weil. mentally naive pigeons came to peck a briefly lighted
response key when keylight offset was followed by
Definition of Free Food response-independent food. This "autoshaping"
Some experimenters refer to intermittently avail- phenomenon has been shown with a number of non-
able food presented on response-independent avian species (see Hearst & J enkins, 1974) and
schedules as free food; however, food availability appears to persist when responses result in reinforcer
is specified by the experimenter. Free food in the omission (Stiers & Silberberg, 1974; Williams &
present review refers to food that is both response- Williams, 1969).
independent and continuouslyavailable. The acquisition and maintenance of responding on
response-independent schedules is not easily
TUEORETICAL IMPLICATIONS accounted for by reinforcement theory (cf. Bindra,
1974; Bolles, 1972), despite the suggestion that
Least Effort responding is superstitiously maintained by adven-
Substantive experimental evidence suggests that titious correlations between behavior and reinforce-
animals tend to maximize reward while simul- ment (Herrnstein, 1966; Zeiler, 1968).
taneously minimizing effort. For example, rats prefer The acquisition and maintenance of food-
the shorter, less effortful path to a goal (DeCamp, acquiring responses when free food is available also
1920; Gengerelli, 1930; Kuo, 1922; Logan, 1960; challenges a reinforcement analysis. If reinforcement
Thompson, 1944; Tolman, 1955; Waters, 1934; strengthens those behaviors which precede it, then
Yoshioka, 1929). More recently, Herrnstein (1958) free food should rein force nonresponding (cf ,
showed that pigeons permitted to choose between Gibbon, Berryman, & Thompson, 1974). Making a
responding for food on different fixed-ratio specific response for food in the presence of free
schedules of reinforcement that were concurrently food appears to be another example of behavior in-
available invariably chose the smaller ratio schedule. adequately accounted for by reinforcement theory.
The least effort hypothesis, as exemplified by these
data, occupies a central position among most RESPONDING FOR REINFORCERS IN TUE
current learning theories (e.g., Ferster & Skinner, PRESENCE OF FREE REINFORCERS
1957; Hull, 1943; Tolman, 1955).
Although some studies have found that animals In the premier experiment (Jensen, 1963), rats were
make the least effortful response (see Lewis, 1965, shaped to leverpress for food pellets. Following 40,
for a review), other studies suggest either that re- 80, 160, 320, 640, or 1,280 continuously reinforced
sponse effort has little effect on choice between responses, a cup of free pellets was placed in the rear
similar rewards or that the more effortful response of the chamber. Jensen found that in one 40-min
is preferred (Collier & Jennings, 1969; Jennings & session the proportion of pellets obtained by lever-
Collier, 1970; Kanarek & Collier, 1973). That pressing increased as a logarithmic function of the
animals respond for reinforcers when abundant free number of reinforced prechoice responses. Rats that
reinforcers are available appears particularly difficult received 40 reinforced responses prior to choice test-
to reconcile with least effort concepts, especially ing responded for about 20070 of their food, whereas
when many responses per reinforcement are emitted rats that received 1,280 reinforced responses before
(e.g., Neuringer, 1970). choice testing responded for 75070 of their food. This
finding was subsequently extended by Neuringer
Response Dependency (1969). After shaping the keypeck response and
The major tenet of reinforcement theory is that receiving seven training sessions on CRF, two
reinforcement increases the probability that those pigeons were housed in chambers where food was
responses which precede it will recur (Skinner, 1938, available either by keypecking or from a cup of free
1948). Arranging to have reinforcers dependent food located in the rear of the chamber. Although
upon a response insures that the response will occur food was continuously available, these animals
in close temporal proximity to reinforcement and emitted approximately 420 and 1,200 responses per
allows the reinforcement process to operate at an day, respectively, for food. A similar finding was
optimal level. One way to assess the functional role reported for two rats, although the average number
of response dependencies is by scheduling rein- of responses was lower than for the pigeons. Condi-
forcers independently of behavior. Differences be- tions were then employed to control for the possible
tween behavior initiated or maintained by schedules reinforcing effects of stimulus feedback associated
of response-dependent and response-independent with response-dependent food presentation. For one
reinforcement can then be attributed to the presence pigeon, a transparent shield was placed across the
or absence of an explicit response-reinforcer depen- feeder opening, for the other the grain hopper wasFREE FOOD 223
emptied, and for the rats, responses produced only that pigeons continued to respond for food on a VI
the auditory feedback provided by the pellet 3-min schedule when free food was available. Al-
dispenser. In all cases, responding decreased to though Neuringer (1970) and Rachlin and Baum
nearly zero. Therefore, the maintenance of respond- (1972) used well-trained pigeons maintained at 80070
ing did not appear attributable to the reinforcing of their free-feeding weights, such conditions are
effect such stimulus feedback might have acquired as not necessary für the maintenance of responding on
a result of pairing with food presentations or to the intermittent schedules of reinforcement. Bilbrey
reinforcing properties of stimulus feedback alone. In et al. (1973) found that pigeons allowed continuous
a second experiment, Neuringer (1969) showed that access to free food, both in the horne cage and during
responding could not be attributed to habit strength experimental sessions, continued to respond for food
of a previously learned response or to special pro- on VI 3-min and VI 30-sec schedules. In a second
perties of the keypeck: Two experimentally naive study, they showed that previous experience with
pigeons lived in their chambers where grain was keypecking for grain had only minor effects on the
dependent on an operant response or available from rate of responding maintained by a VI 30-sec
a cup of free food. For one pigeon, the response schedule.
was the depression of alever, for the other a key- Sawish and Denny (1973) also reinforced pigeons'
peck. In both instances, pigeons learned, without keypecks for food on a VI I-min schedule, but they
prior shaping, to respond for food and continued to made the opportunity to respond for food dependent
respond despite the availability of free food. on eating or not eating free food. They found that
It also has been shown that rats continuously eating free food increased when the opportunity to
housed in achamber with free food available readily respond was dependent on eating free food and de-
acquire and maintain a food-producing operant re- creased when the opportunity to respond was depen-
sponse (Coburn & Tarte, 1976; Kopp et al., 1976; dent on not eating free food for aperiod of 30 or
McLaughlin et al. , 1973). Similarly, Baeninger and 60 sec. Sawisch and Denny suggested their data were
Mattleman (1973) found that Siamese fighting fish evidence of the operation of the Premack principle
acquired and continued to emit an operant response (cf. Premack, 1959) in a free food situation; that is,
that produced a mirror to which the fish could the use of a high-probability response (keypecking)
display, although a free mirror was continuously to reinforce a low-probability response (eating free
present. Therefore, responding for reinforcers in the food).
presence of free reinforcers includes both the acquisi-
tion and the maintenance of behavior. Preference for Free Food
Despite the apparent robustness of the phenomenon
Schedule of Reinforcement and the suggestion of earlier research that animals
In the Jensen and Neuringer studies, responses prefer response-dependent food, subsequent studies
were continuously reinforced. Responding also has produced conflicting data. In one experiment
been maintained on intermittent schedules of re- (Taylor, 1972), choice testing was preceded by 1,000
inforcement when free food was concurrently avail- continuously reinforced barpresses and then a single
able. Carder and Berkowitz (1970) preceded choice session where only free food was available. During
testing with three sessions of free food training and the initial test sessions of 20 min, rats responded for
then six sessions of barpress training. They found over 50070 of their food; however, by the 15th test
that when only 1 or 2 responses were required per session, preference for response-dependent food
reinforcement, rats responded for 80070-100070 of decreased to 20070. Taylor suggested that the differ-
their food. However, when the response requirement ence between his and Jensen's data was that Jensen
was increased to 10, preference for response- conducted only one session of preference testing
dependent food decreased to 10070-30%. Tarte and and hence failed to observe the changes toward free
Vernon (1974) investigated barpressing for food food that occur over consecutive sessions. An
across a greater range of fixed-ratio schedules and alternative explanation for the inconsistency in
found a linear decrease in preference for response- preference is that Taylor used sessions of shorter
dependent food as the ratio was increased from FR 1 length. Tarte and Snyder (1973) used a procedure
(65070) to FR 21 (11070). similar to that of Taylor, but conducted 10
Free food responding also has been demonstrated l-h sessions of preference testing. They found a
on aperiodic schedules of reinforcement. Neuringer consistent preference for response-dependent food
(1970) found that pigeons emitted over 40 keypecks with no systematic change in preference evidenced
per reinforcement when keypecks were reinforced across consecutive sessions. That Taylor allowed
following irregular time intervals that averaged 1 min prechoice training with free food also may have
(VI I-min schedule) despite the availability of free affected preferences (see Food Training Procedures
food. Similarly, Rachlin and Baum (1972) showed below). Nonetheless, other studies support the find-224 OSBORNE
ing that animals prefer free food. For example, two fairly constant across sessions. By the third choice
direct replications of the Carder and Berkowitz session there was a clear difference between the
(1970) study showed that rats preferred free over groups: animals in the food-reinforcer group bar-
response-dependent food when responses were re- pressed for 14070 (20-mg pellets) and 40070 (45-mg
inforced by FR 1, FR 2, and FR 10 schedules (Atnip pellets) of their reinforcers; rats in the water-
& Hothersall, 1973; Hothersall et al., 1973). Koffer reinforcer group barpressed for 4070 (0.01 cc) and
and Coulson (1971) also found that cats, given a 31070 (0.1 cc) of their reinforcers. Thus the data of
choice between responding for fish or obtaining it Robertson and Anderson suggest that reinforcer
freely, consistently chose free fish, and Lambe and magnitude exerts a strong effect on preference.
Guy (1973) found that both rats and gerbils pre- Moreover, it appears that there is a greater prefer-
ferred free over response-dependent food. ence for response-dependent food, relative to
response-dependent water reinforcers, even when the
Reinforcer-Motivational Variables magnitude of the water reinforcer (by weight) is twice
The ability of qualitatively different reinforcers that of food; that is, 0.1 cc of water represents a
to maintain responding in the presence of equivalent, larger magnitude reinforcer than does a 45-mg food
free reinforcers also has been studied. Choice pro- pellet, yet under these magnitude conditions rats bar-
portions vary between experiments, but. it has been pressed for a greater proportion of their food
demonstrated that animals show a stronger prefer- reinforcers.
ence for response-dependent food reinforcers than Variation in the magnitude of water reinforcers
for response-dependent water reinforcers under may account for some of the variability in previously
conditions of continuous reinforcement (Carder, reported choice data. For example, Powell (1974)
1972; Roberts & Anderson, 1975; Tarte et al., 1974; and Taylor (1972, 1975) both reported that rats bar-
Taylor, 1972), and intermittent reinforcement (cf. pressed for 5070 or less of their water reinforcers, but
Tarte & Vernon, 1974; Taylor, 1975). Carder (1972) they used response-dependent reinforcers of rela-
exposed rats to a situation where either water or tively small magnitude (0.01-0.02 cc). Alternatively,
sucrose solution reinforcers were available freely Carder (1972) and Tarte et al. (1974) found that rats
or by barpressing. He found that rats obtained 83070 barpressed for 26070 and 54070, respectively, of their
of their sucrose reinforcers but only 23070 of their water reinforcers when the magnitude of the response-
water reinforcers by barpressing. Similarly, Tarte dependent reinforcer was 0.1 cc. Knutson and
et al. (1974) reported preferences of 79070 and 54070, Carlson (1973) reported that rats barpressed more
respectively, for response-dependent food and for water reinforcers than for food reinforcers when
response-dependent water reinforcers. equivalent free reinforcers were present. They used
Altering the physical quality of the reinforcer also standard 45-mg food pellets and 0.01 cc of water as
affects preference. Carder (1972) found that when the response-dependent reinforcers. Therefore, al-
both free and response-dependent reinforcers were though rats responded more for water reinforcers,
adulterated with quinine, preference for response- they still may have obtained a larger proportion of
dependent sucrose decreased as a function of adulter- their food reinforcers by responding (cf. Robertson
ation level. Tarte et al. (1974) found that a non- & Anderson, 1975).
nutritive reinforcer (saccharin) maintained little In sum, preference for response-dependent rein-
responding with either food- or water-deprived rats forcements appears to be an increasing function of
when saccharin was freely available, and Davidson the effectiveness of the reinforcer, which depends
and Davis (1975) found that the proportion of food upon reinforcer-motivational variables (cf', Bindra,
acquired by barpressing decreased following in- 1974), such as rate and magnitude of reinforcement
jections of known anorexic drugs. as well as the physical quality of the reinforcer. There
Robertson and Anderson (1975) investigated the also is a greater preference for response-dependent
effects of varying the magnitude of the response- food compared to response-dependent water rein-
dependent reinforcer using both food (20- or 45-mg forcers when equivalent free reinforcers are avail-
pellets) and water (0.01 or 0.1 cc) reinforcers. They able. This latter finding is consistent with other re-
reported that differences in choice proportions across search (e.g., Fallon, Thompson, & Schild, 1965)
reinforcer magnitudes were not statistically signifi- showing that water is a less potent reinforcer than
cant. However, no prechoice exposure was provided food when deprivation duration and reinforcer
to the free reinforcers and all three choice sessions weight are equated.
were included in their data analysis. A close examina- An indirect way to manipulate reinforcer effective-
tion of the data shows that preference for small- ness is by varying the severity of reinforcer depriva-
magnitude response-dependent reinforcers decreased tion. Given the above data, food deprivation and
linearly over sessions, whereas preference for large- preference for response-dependent food might be
magnitude response-dependent reinforcers remained expected to covary. In the single experiment in whichFREE FOOD 225
food deprivation was varied systematically (Tarte reinforced responses (Jensen, 1963). McLaughlin
& Synder, 1972), preference for response-dependent et al. (1973) found that rats left in experimental
food increased as food deprivation time was extend- chambers für 7 to 10 days acquired a barpress
ed from 0 to 92 h. Morgan, Einon, and Nicholas response with both free food and water contin-
(1975) also found that rats deprived of food for 23 h uously available. Although response rates initially
responded for a significantly greater proportion of were low, removal of the free reinforcers for 24 h
their food than rats fed immediately before a session. increased barpressing tenfold once a choice between
However, other researchers failed to detect similar free and response-dependent food or water was rein-
deprivation effects.' Duncan and Hughes (1972) stated. In a more systematic investigation of the
found no difference in the proportion of food effects of prechoice training, Tarte and Synder (1973)
chickens obtained by keypecking between conditions found that rats barpressed for 72070 of their food
of continuous access to food and 22-h food depriva- following three free food sessions and six sessions
tion. Davidson (1971) trained rats to keypress of barpress training, but they barpressed for only
for food on a multiple schedule of reinforcement. 15070-20070 of their food following equalized food
When the chamber was illuminated, every 10th re- training procedures.
sponse produced reinforcement (FR 10) and ex- Mitchell et al. (1973) suggested that the prechoice
tinguished the chamber illumination, signaling that training schedule, per se, is a less important variable
response-dependent food was not available. Cessa- than the amount and/or recency of familiarity with
tion of responding for 20 sec (DRL 20 sec) reinstated the different food sources. After shaping, rats were
the chamber illumination and the FR 10 schedule. allowed to make 140 continuously reinforced re-
When free food was provided, it was consumed dur- sponses and were then given 10 training sessions. One
ing the DRL component, but rarely during the FR group received seven sessions of barpress training
component. Providing continuous access to food in followed by three sessions of free food training; a
the horne cage eliminated within-session consump- second group received the same distribution of train-
tion of free food but failed to abolish keypressing ing sessions, but in reverse order. When choice test-
in the FR component, ing immediately followed barpress training rats bar-
Several investigators (Davidson, 1971; Knutson & pressed for 50070 of their food, but they barpressed
Carlson, 1973; MitchelI, Scott, & Williams, 1973; for only 1070 of their food when choice testing im-
Robertson & Anderson, 1975) noted that the prob- mediately followed free food training. A third group
ability of a response within an experimental session of rats given three alternating free and response-
increases over time; free food typically is consumed dependent food training sessions followed by four
during the first part of a session. It appears, then, sessions of barpress training responded for 34070 of
that probability of responding and within-session their food. Thus preference for free food appeared to
food deprivation are inversely related. It remains be an increasing function of the recency of free food
possible that within- and between-session food training. However, free food training sessions were
deprivation have different effects on responding. conducted with the response bar present but in-
Variables that control cessation of responding for operable. Consequently, barpressing for food may
food (or eating) are poorly understood (Morgan, have undergone extinction during free food training
1974a). At present the effects of food deprivation on and the decreased preference for response-dependent
preference for response-dependent food are food rnay have been due to the recency of
equivocal. However, variables other than "hunger extinction.
drive" or food deprivation appear responsible for the There also is some evidence that familiarity with
maintenance of free food responding, since several the food containers (Mitchell et al., 1973; MitchelI,
reports have shown that animals may fail to consume Williams & Sutter, 1974; but see Carder & Beckman,
up to 25070 of the food they have responded for 1975) or the food substance (Carlson & Riccio, 1976)
(Mitchell et al. , 1973; Neuringer, 1969; Tarte & affects preference. Mitchell et al. (1973) found that
Synder, 1973) and that considerable responding is rats barpressed for 48070 of their food when free
maintained in the absence of food deprivation (e.g., food was placed in a novel container and response
Kopp et al., 1976; Neuringer, 1969; Wallace et al., dependent food was presented in a familiar con-
1973). tainer, but they barpressed for only 1070 of their food
when conditions were reversed. Carlson and Riccio
Food Training Procedure (1976) found that providing prechoice exposure to
Providing differential prechoice training with food pellets of the type used during experimental
response-dependent food significantly increases pre- sessions, either in the horne cage or in the test
ference for response-dependent food. As previously chamber, reduced subsequent preferences for
noted, preference for response-dependent Iood in- response-dependent food. In a second experiment, it
creases as a function of the number of prechoice was shown that changing the flavor of both free and226 OSBORNE response-dependent pellets between conditions of sponse of the rat. Wild rats tend to avoid strange prechoiee training and testing decreased preferences food and strange food sources (Barnett, 1963); for response-dependent food. laboratory rats apparently share this disposition. Stolz and Lott (1964) investigated the effects of That rats run over a pile of pellets in a straight alley prechoiee training on a running response. Rats were (Stolz & Lott, 1964) seems explicable by such an trained to run an 8-ft straight alley for a single food analysis. Rats prefer to eat in that part of the pellet; then, during choiee trials, a pile of 1,500 apparatus where they have been accustomed to pellets was placed midway down the alley. Rats that eating; for example, they will carry food found in the received more prechoice training more often ran to stern of a maze to the goalbox before consuming the goalbox than animals given less training. How- it (Cohen-Salmon & Blancheteau, 1967). On the ever, two subsequent studies suggest that providing other hand, Carlson and Riecio (1976) argued that prechoiee training for a running response increases during barpress training the reinforcer becomes a preference for free food. Leung, Jensen, and Tapley discriminative stimulus for the operant response (1968) allowed two groups of rats either 75 or 285 that produces it (cf. Denny, 1971), and thereby main- reinforced runs down a straight alley before giving tains barpressing in the presence of free food. Yet them a choiee between eating free food in the start- another interpretation of the effects of prechoiee box or running the alley for a single pellet. They food training was provided by Taylor (1975), who found that rats given fewer prechoice trials ate less suggested that preference for response-dependent free food and spent less time in the startbox before reinforcers results from a failure to discriminate be- traversing the alley. Jensen, Leung, and Hess tween conditions of testing and training. He found (1970), using a similar procedure, allowed three that preference for response-dependent water rein- groups of rats 0, 40, or 285 reinforced runs before forcers decreased whenever the response requirement choiee testing. They found that rats given no pre- was increased or decreased from its prechoiee train- choiee training spent less time eating free food before ing value. The wide range of species, reinforcers, and running the alley than rats given 40 or 285 prechoiee procedures used in free "food" studies makes it trials; there was no signifieant difference between difficult to ascribe the effects of training procedures the latter two groups. to food source familiarity or discrimination effects. Free food experiments typically invo1ve a choice In many cases there is little to distinguish between between topographically dissimilar operants that such accounts because common predietions are often lead to the same reinforcer. Larson and Tarte (1976) made by all. However, the discriminative stimulus assessed the effects of prechoiee training when the hypothesis is consistent with data collected outside choiee responses involved topographically similar the free food area (cf, Rescorla & Skucy, 1969) and operants. They found that rats initially preferred the it can more easily account for the effects of pre- longer of two paths to a goalbox in a T-maze if they choiee training when only a single food source is were given more prechoice training with the longer availab1e (e.g., Kleinman et al., 1976). Nevertheless, path. However, over successive choice trials, prefer- of the studies reviewed, an explicit preference for ence for the shorter path increased to about 90070. response-dependent reinforcement was obtained only Rats given an equal number of prechoice trials for when no prechoice training for the free reinforcer both paths preferred the shorter path from the out- was provided (e.g., Jensen, 1963) or when prechoice set of testing. Kleinman, McLaugh1in, Gerard, training procedures that favored responding were Bosza, and Clipper (1976) found that rats consistent- used (e.g., Carder & Berkowitz, 1970). Together, ly preferred the larger of concurrently available FR 1 these data help resolve the preference issue: animals and FR 3 schedu1es following differential training appear to prefer response-dependent reinforcers with the larger ratio schedule. Taken together these only after asymmetrieal food training procedures. two studies suggest that prechoiee training effects are However, the issue of why animals respond at all transient if the difference in "effortfulness" be- when free reinforcers are available is left unresolved. tween the two alternatives is large (e.g., running a path of 122 or 244 cm; Larson & Tarte, 1976) Stimulus Change compared to the more enduring effects that resu1t Several recent experiments suggest that free food when the difference in effortfulness is small (e.g., responding is maintained by stimulus changes cor- barpressing on an FR 1 or FR 3 schedule or, per- related with response-dependent food presentation. haps, barpressing on an FR 1 schedu1e for pellets or Wallace et al. (1973) provided naive pigeons with a obtaining them freely). choiee between two sources of free food. They found Several explanations have been proposed for the that when entries into one food source were accom- effects of prechoice food training. Mitchell and his panied by a change in ambient stimulus conditions, associates suggested that the effect of food source pigeons consistently preferred the food source associ- familiarity on preference is due to a neophobic re- ated with stimulus change. In a second experiment,
FREE FOOD 227
two pigeons lived in experimental chambers where Two explanations have been proposed for the
food was avilable by keypecking or from a cup of effects of stimulus change in free food experiments.
free food. Whenever keypecks produced both food One suggests that stimulus change associated with
and stimulus change associated with food presenta- response-dependent food serves as a conditioned
tion, more than 100 continuously reinforced re- reinforcer; the other maintains that stimulus change
sponses were made per session. However, respond- derives its reinforcing properties through sensory
ing decreased to nearly 0 if stimulus change was reinforcement effects.
presented contemporaneously with free food acquisi-
tion, and responses produced food without con- Conditioned Reinforcement
comitant stimulus change. Alferink et al. (1973) Alferink et al. (1973) and Tarte and Synder (1972)
trained pigeons to respond for grain on a FR 300 suggested a conditioned reinforcement explanation
schedule and then propped up the grain hopper so of the free food phenomenon. Conditioned rein-
that grain was available without responding. Re- forcers are stimuli that acquire their reinforcing
sponding was maintained when every 300 responses property through association with an unconditioned
continued to produce the hopper light, but decreased reinforcer. Two sources of stimulus change associ-
to a low level when responses had no scheduled ated with food presentation in free food experiments
effect. Reinstating the hopper-light-on contingency are those afforded by the response operandum and
returned responding to its previous rate. Responding the food delivery system, and any scheduled stimu-
appeared to be controlled by the response-dependent lus changes. Through repeated pairing with food
presentation of a stimulus (hopper light) previously presentation, stimulus change may acquire condi-
associated with response-dependent food. tioned reinforcing properties. Responding for food
Osborne and Shelby (1975) extended these findings in the presence of free food might be attributed to
to a nonavian species. Following six alternating free the combined reinforcing effectiveness of response-
and response-dependent food training sessions, rats produced food and conditioned reinforcement
were provided a choice between barpressing for provided by stimulus change. Neither response-
pellets on CRF or obtaining them freely from a cup produced food alone (Osborne & Shelby, 1975;
of 500 pellets. When abrief auditory and visual Wallace et al., 1973) nor resonse-produced stimulus
stimulus accompanied barpresses, these animals change alone (Davidson, 1971; Duncan & Hughes,
responded for 28070 of their food; however, removal 1972; Enkema, Slavin, Spaeth, & Neuringer, 1972;
of the stimulus as a consequence of barpressing de- Neuringer, 1969; Osborne & Shelby, 1975; Tarte,
creased responding to nearly zero. Subsequent condi- 1974) are sufficient to maintain responding when free
tions revealed that the amount of maintained re- food is available. Yet, when responses produce both
sponding, and consequently preference for response- food and stimulus change, many responses are
dependent food, varied as a function of stimulus emitted. The failure to maintain responding with
change conditions attendant upon response-dependent stimulus change alone is consistent with the finding
food presentation. For example, there was a signifi- that stimuli which serve as conditioned reinforcers
cantly greater preference for response-dependent often lose their effectiveness when primary reinforce-
food when barpresses were accompanied by both ment is withheld (Kelleher & Gollub, 1962).
auditory and visual feedback than when barpresses
were accompanied by auditory or visual feedback Sensor Reinforcement
alone. Response-dependent sensory reinforcers unrelated
A fortuitous example of the effects of stimulus to primary appetitive reinforcers are reinforcing
change was provided by Mitchell et al. (1973). They (cf. Kish, 1966). Naive pigeons acquire and main-
allowed one group of rats a choice between bar- tain a keypeck response when the sole consequence
pressing for pellets or obtaining them freely from a of responding is a change in ambient stimulus condi-
pile of 300 pellets in an adjacent food dish. To con- tions or a change in response key color (Appel, 1963;
trol for the possible novel effect of bulk reinforce- Herrnstein & Loveland, 1972). Osborne and Shelby
ment (300 pellets), a second group of rats received (1975) suggested that animals in free food experi-
a food dish that contained only 20 free pellets; an ments respond for food plus stimulus change not
additional pellet was dispensed by the experimenter because stimulus change is a conditioned reinforcer,
each time a free pellet was removed. Although both but because the reinforcing value of stimulus change
groups preferred free food, rats provided with free increases in the presence of other more potent
food accompanied by the sound of the pellet dis- primary reinforcers. Sensory reinforcement effects
penser consumed less free food than rats provided may interact with the primary reinforcement effects
free food without the sound of the feeder. In short, of response-dependent food such that the total rein-
equating stimulus change conditions for both free forcement obtained by responding is sufficient to
and response-dependent food essentially eliminated maintain responding when free food is available.
barpressing. In all the free food experiments reviewed here,228 OSBORNE stimulus change was paired with food presentation. different environments. Coburn and Tarte (1976) This makes it difficult to distinguish between condi- reared groups of rats in either stimulus-impoverished tioned and sensory reinforcement effects, because or stimulus-enriched environments and then housed many of the variables that affect the reinforcing them in chambers where both free and response- effectiveness of conditioned reinforcers have a dependent food were avilable. Animals in the im- similar effect on the reinforcing potential of sensory poverished group readily acquired the barpress re- reinforcers. However, some free food data appear sponse and consistently responded for ab out 600;0 of more amenable to a sensory reinforcement analysis. their daily food intake; rats in the enriched group First, if conditioned reinforcers acquire their rein- responded at a low level throughout the experiment forcing properties through repeated association with and earned less than 1010 of their food by responding. primary reinforcers, then acquisition of a food- Morgan et al. (1975) assessed choice behavior of obtaining response in the presence of free food (Kopp impoverished and enriched rats on a concurrent et al. , 1976; Neuringer, 1969) is difficult to explain schedule of reinforcement. On one side of a by conditioned reinforcing effects of stimulus change. chamber, leverpresses produced a food pellet once On the other hand, sensory, or unconditioned every 30 sec; on the other side of the chamber, a reinforcing stimuli, need no prior exposure to affect pellet was delivered once every 30 sec, independently behavior and they presumably would be reinforcing of the animal's behavior. In this situation there was from the outset. Second, Tarte and Synder (1973) no difference in the choice behavior of impoverished required that 0, 25, or 50 continuously reinforced and enriched rats. When these same rats were sub- responses occur prior to placing free food in the sequently provided a choice between free food and chamber. During the ensuing choice period, respond- leverpressing for pellets on a schedule of continuous ing decreased as the number of reinforced prechoice reinforcement, the impoverished rats responded for responses increased. Although the change was not 42010 of their food whereas animals in the enriched statistically significant, it is noteworthy that this group responded for only 19010 of their food. In the requirement should have increased the conditioned first experiment, identical stimulus change ac- reinforcing strength of stimulus change because of companied both response-dependent and response- the extra number of stimulus-food pairings; yet independent pellet deliveries, but in the second preference for response-dependent food decreased. experiment, stimulus change only accompanied Alternatively, with such a procedure, sensory rein- response-dependent pellets. Therefore, differences forcement effects might have undergone partial satia- in preference behavior between enriched and im- tion. Finally, following a condition where no poverished animals were evident only when differ- stimulus feedback accompanied responding, Osborne ential stimulus change accompanied response- and Shelby (1975) exposed some rats to a condition produced pellets. where an auditory stimulus accompanied res pond- Davis, Beighley, Libretto, Mollenhour, and ing; for others a visual stimulus accompanied re- Prytula (1975) also found that rats reared in stimulus- sponding. Addition of the auditory stimulus in- deprived environments preferred response-dependent creased responding, although this stimulus previous- food, whereas rats reared in either normal or stimulus- ly had not been associated with food presentation. enriched environments preferred free food (but see Subsequently, when both auditory and visual feed- Tarte, Townsend, & Vernon, 1973). back accompanied response-produced food, re- Numerous behavioral differences exist between sponding increased above the level maintained by animals reared in impoverished and enriched en- either auditory or visual feedback alone. Egger and vironments (cf. Morgan, 1973, Morgan et al., 1975; Miller (1962, 1963) have shown that, when stimuli Wallace, 1974). For example, impoverished rats provide redundant information ab out impending show heightened levels of exploratory behavior and reinforcement, such stimuli do not become condi- are slower to habituate to environmental (sensory) tioned reinforcers even though they are temporally stimuli. It might be argued, then, that impoverished contiguous with reinforcement. Adding stimulus rats are more likely to respond for stimulus change feedback from a different stimulus modality, as or satiate more slowly than normals. This interpreta- Osborne and Shelby did, added no information to tion is not only consistent with the data of Coburn the situation, yet responding increased. This squares and Tarte (1976) and Morgan et al. (1975), but also with the fact that increasing the complexity, the is supported by a large body of literature which inconstancy, or the novelty of a sensory reinforcer in- shows that the ability of sensory reinforcers to main- creases its reinforcing potential (Barnes & Baron, tain responding is directly related to antedating 1961; Dember, 1956; Welker & King, 1962). stimulus deprivation conditions (cf. Fowler, 1971; Sensory reinforcement also may account for some Kish,1966). of the differences in acquisition and maintenance of Although a sensory reinforcement explanation free-food responding between animals reared in of stimulus change appears to account for much of
FREE FOOD 229
the data, it does not account for the effects of pre- to encompass existing data collected within the free
choice training on preference. Here a conditioned food paradigm: The tendency to respond for rein-
reinforcement explanation is more consistent; that is, forcers when physically equivalent free reinforcers
the strength of a conditioned reinforcing stimulus are available (a) increases whenever reinforcer
presumably increases with repeated stimulus- effectiveness increases, and/or when the reinforcing
reinforcer pairings. effectiveness of the stimulus change associated with
its presentation increases, and (b) decreases when-
Controlling Variables ever the reinforcing effectiveness of the reinforcer or
Prechoice food training appears to affect prefer- the stimulus change associated with its presentation
ence between free and response-dependent food. decreases. The reinforcing effectiveness of primary
However, differential prechoice training with reinforcers like food and water is determined by a
response-dependent food is neither necessary nor host of variables that include the rate and magnitude
sufficient for the acquisition or maintenance of of reinforcement, the physical quality of the rein-
responding for reinforcers in the presence of forcer, and reinforcer deprivation. Several variables
equivalent free reinforcers. Although food source that contribute to the reinforcing effectiveness of
familiarity or discriminative stimulus hypotheses sensory reinforcers include the complexity, novelty,
may aid our understanding of the effects of prechoice and inconstancy of the stimulus change as well as
training on preference, neither adequately accounts antedating stimulus deprivation conditions. Al-
for the wide range of data reviewed here. First, though the emphasis here is on the sensory rein-
differential prechoice training with response- forcing properties of stimulus change, a comparable
dependent food does not insure preference for argument can be made for the conditioned rein-
response-dependent food (Atnip & Hothersall, 1973; forcing properties of stimulus change, i.e., the
Hothersall et al. , 1973). Second, both rats and strength of a conditioned reinforcing stimulus pre-
pigeons readily acquire an operant response for food sumably covaries with the reinforcing effectiveness
or water reinforcers in the presence of equivalent of the primary reinforcer with which it is associated.
free reinforcers (Kopp et al., 1976; McLaughlin et al., It remains possible that stimulus change serves both
1973; Neuringer, 1969), and continue to respond for a conditioned and a sensory reinforcement function.
reinforcers after equalized training procedures
(Osborne & Shelby, 1975; Tarte & Synder, 1973).
Finally, prechoice training appears to have only a Interexperiment Variability
minor effect on the rate of maintained responding Considerable interexperiment variability has
in avian species (Bilbrey et al., 1973; see also Duncan occurred in studies of preference between free and
& Hughes, 1972). Alternatively, responding fails to response-dependent food. Much confusion and
develop unless differential stimulus change difficulty in integrating free food data into existing
accompanies response-dependent food presentation theory has resulted. Methodological differences are
(Mitchell et al. , 1973); well-trained responses are an obvious source of variance, although the problem
extinguished readily if the stimulus change that is complicated further by the use of different de-
normally accompanies response-dependent food pendent measures. Data have been reported as
presentation is withheld (Alterink et al. , 1973; number of responses (Neuringer , 1969), propor-
Wallace et al., 1973); and preference for response- tional amounts of free and response-dependent food
dependent food appears to vary systematically as a consumed (Carder & Berkowitz, 1970), and latency
function of the stimulus change conditions that to eat free or response-dependent food (Jensen et al.,
accompany responses. Despite the present difficulty 1970; Koffer & Coulson, 1971). Nonetheless, inter-
in ascribing a conditioned or sensory reinforcement experimentvariability in preference data appear to be
function to stimulus change, it seems clear that attributable largely to differences in prechoice food
responding for reinforcers when physically equiv- training procedures and to stimulus change condi-
alent reinforcers are freely available is attributable to tions associated with response-dependent food, al-
the differential reinforcement afforded by response- though session duration differences and the brevity
dependent food and attendant stimulus change. of choice testing may also contribute to variability.
However, this appears to be an interactive effect; Differences in stimulus change conditions between
that is, the amount of responding maintained by experiments may be no greater than those resulting
food-plus-stimulus change cannot be predicted by a from apparatus differences. Such differences may
simple additive combination of the amount of nonetheless affect choice behavior and thereby pro-
responding maintained by each (Os borne & Shelby, vide a possible explanation for differences in ob-
1975; cf. Herrnstein & Loveland, 1972). tained preferences between experiments that used
In considering the interactive effect of food and seemingly identical procedures (cf'. Atnip &
stimulus change, several general staternents appear Hothersall, 1973; Carder & Berkowitz, 1970).230 OSBORNE
RELATEDRESEARCH are indifferent between response-independent and
response-dependent schedules of reinforcement when
Response-Independent and Response- the rate of reinforcement is constant for each; devia-
Dependent Reinforcement tions from indifference appear to be attributable to
Studies by Morgan (1974b) and Singh (1970) asymmetrical stimulus conditions. That animals
typically have been included in evaluations of the free show a near-exclusive preference for free over
food phenomenon. Because "free food" was not response-dependent food when stimulus change
continuously available in these studies, they differ conditions are equalized may result from differ-
from the procedural arrangement used in other free ences in rate of reinforcement.
food studies. Singh (1970) allowed rats a choice be-
tween barpressing for pellets on one side of a Signaled Reinforcement
chamber or receiving free pellets on the other side. Animals and humans permitted to choose between
"Free pellets" were dispensed only if the animal a primary reinforcer preceded by a signal and the
remained in the free food compartment for a time same reinforcer unsignaled prefer the signaled rein-
inter val equated with the average obtained inter- forcer (Cantor, 1971; Cantor & LoLordo, 1970,
reinforcement interval of concurrently available 1972; Furedy & Klajner, 1972; Lewis, Lewin,
response-dependent food. When barpresses were Muehliesen, & Stoyak, 1974; but see Hershiser &
reinforced on fixed-ratio schedules (FR 1, FR 3, and Trapold, 1971). Cantor and LoLordo (1970) ad-
FR 11), rats responded for about 70070 of their food, ministered 0.5-sec trains of electrical stimulation of
and they continued to prefer response-dependent the brain (ESB) to rats according to a variable-time
food even when free food was available at a faster I-min schedule. The rats were placed in a shuttlebox
rate. and provided a choice between receiving ESB pre-
The finding of a consistent preference for response- ceded by a signal on one side of the chamber and
dependent food appears inconsistent with the stimu- the simultaneous presentation of ESB and signal
lus change hypothesis outlined above inasmuch as on the other. They found that rats spent approxi-
stimulus change presumably was equalized for both mately 70070 of their time in that part of the chamber
food sources. However, free pellets could arrive where ESB was preceded by a signal. Lewis, Lewin,
when the rat was anywhere in the free food compart- Muehliesen, and Stoyak (1974) reinforced pigeons'
ment, whereas on the response-dependent side the keypecks with grain on a VI l-min schedule. A peck
rat had to be in contact with the response lever and to a second, concurrently available changeover key
in close physical proximity to the food cup at time of produced a VI l-min schedule for 1 min, during
pellet delivery. Consequently, the occurrence of which reinforcement was preceded by a 5-sec tone
stimulus change and food acquisition could be separ- stimulus. Under this procedure, pigeons pecked the
ated temporally on the reponse-independent side but changeover key at a rate sufficient to maintain the
presumably would occur in close temporal proximity signaled schedule in effect 90070 of the time. In
on the response-dependent side. Within a similar studies of signaled reinforcement, a stimulus pre-
paradigmatic arrangement, Morgan (1974b) found cedes and is invariantly correlated with reinforce-
that food consumption latencies for response- ment. In free food experiments, a response is
independent food were much higher than for followed by a brief stimulus and food. Despite pro-
response-dependent food. Morgau's data suggest cedural differences between the two paradigms, the
that the temporal relationship between stimulus stimulus effects on behavior are similar. Differ-
change and food acquisition may be an important entially adding a stimulus that precedes reinforce-
variable. When Morgan equated stimulus change ment in a choice situation increases preference for
conditions attendant upon food acquisition, he that alternative.
found that rats responded for only 30070-40070 of
their food. The obtained rate of reinforcement Autosbaping and Automaintenance
(number of pellets/time) was higher on the response- Experimentally naive animals "autoshape" to
dependent side and yielded a close approximation stimuli that signal response-independent reinforce-
to the "matching law" (Herrnstein, 1970). When an ment (e.g., Brown & Jenkins, 1968; Peterson, Ackil,
added stimulus signaled that a barpress would be Frommer, & Hearst, 1972) and still respond when
reinforced, rate of reinforcement and preference responses prevent the delivery of reinforcement
for that side increased. When rate of reinforcement (Stiers & Silberberg, 1974; Williams & Williams,
and stimulus conditions were equalized for both 1969). Autoshaping and free food experiments differ
food sources, an equivalent amount of food was in terms of alternative sources of reinforcement, but
obtained from each. In short, the Singh and Morgan are similar because responding in both cases is
data are consistent with other research (Killeen, 1968; initiated and maintained in a situation where rein-
Moore & Fantino, 1975) which indicates that animals forcers are available independently of behavior.FREE FOOD 231
In arecent monograph, Hearst and Jenkins (1974) reinforcement may explain preference for signaled
suggested that animals orient towards, approach, over unsignaled schedules of reinforcement, it has
and make contact with stimuli that signal impending difficulty explaining some free food data and appears
reinforcement and avoid or withdraw from stimuli to be an unsuitable explanation for autoshaping and
that signal reinforcer unavailability, a behavioral automaintenance data (cf. Hearst & Jenkins, 1974;
process they call "sign-tracking." Autoshaping and but see Hursh, Navarick, & Fantino, 1974; Zentall
related phenomena are attributed to stirnulus- & Hogan, 1975). Presently, a single unifying account
reinforcer effects; i.e., stimuli that precede reinforce- that encompasses the range of stimulus-reinforcer
ment elicit approach and contact behavior to those effects reviewed here is lacking, but future research
stimuli. Unconditioned stimulus events also appear on sign-tracking behavior may fill this void. A sign-
to elicit exploratory reactions of approach, investiga- tracking analysis of the free food phenomenon is not
tion, and contact behavior (cf. Fowler, 1971). Such currently possible because of the paucity of data on
"sensory reinforcers" appear to have an effect on the effects of systematically varying stimulus condi-
the initiation and maintenance of responding when tions correlated with response-dependent food. This
free food is available, and they may contribute to the deficit itself suggests a productive avenue for future
initiation and maintenance of autoshaped responses. free food research. Data forthcoming from such
Herrnstein and Loveland (1972) found that naive research not only may allow closure on the free food
pigeons autoshaped to a stimulus change reinforcer. phenomenon, but may provide data relevant for
They concluded that "this kind of reinforcement other areas of research in animallearning as well.
(stimulus change) would by itself, make possible the
phenomenon of autoshaping." However, the SUMMARY
phenomenon is greatly enhanced by correlating food
with stimulus change. Similarly, Blanchard and A plethora of theoretical accounts have been
Honig (1976) and Downing and Neuringer (1976) adduced to explain the free food phenomenon. Sorne
found that naive pigeons pecked an illuminated of these include: an intrinsic appeal of performing
response key the first time it was presented and be- the operant (Jensen, 1963), responses serving as their
fore pairing with grain if sufficient hopper training own motivation and reinforcement (Neuringer, 1969;
was provided. Kopp & Bourland, Note 1), discriminative stimulus
Not all stimuli have an equiva1ent effect on be- properties of reinforcers (Carlson & Riccio, 1976),
havior, either prior to (Kish, 1966) or after the con- environmental competence (Singh, 1970; Stephens,
ditioning process (Revusky & Garcia, 1970; Seligman, Metze, & Craig, 1975), arousal properties of rein-
1970). Stimulus events therefore may derive their forcers (Carder, 1972), deprivation (Tarte & Synder,
reinforcing effectiveness collectively from both con- 1972), habit strength or behavioral persistence
ditioned and unconditioned properties (cf. Osborne (Morgan, 1974a; Tarte & Synder, 1973), neophobia
& Shelby, 1975). (Mitchell et al., 1973), Premack's principle (Sawisch
& Denny, 1973), and discrimination (Taylor, 1975).
Self-Reinforcement While all of these theories may account for part of
Mahoney and Bandura (1972) found that pigeons the data, none appears to encompass the broad range
trained to peck a briefly illuminated response key of experimental findings reviewed here.
before eating continued to respond when the response- It has perhaps been the mistaken belief that free
dependency for food was removed. They interpreted and response-dependent food are equally attractive
their data as an anima1 analogue of human self- (reinforcing) that has engendered both surprise and
reinforcement (Mahoney, 1972). However, respond- interest in the free food phenomenon. Results of
ing in Mahoney and Bandura's experiment was main- these studies have tempted some (Singh, 1972b) to
tained only when responses were accompanied by draw analogies between the propensity of animals
a brief stimulus change. Consequently, responding to work for reinforcers in the presence of free rein-
in this case appears similar to responding main- forcers and the human welfare system. Others may
tained in free food experiments; i.e., animals appear have construed such results as an indictment of
to prefer food-plus-stimulus change over food alone. current learning theories, As shown by the data re-
The data reviewed here show that stimulus- viewed above, however, the free food anomaly may
reinforcer relationships have a strong effect on be- be more in perception than in reality; it is the
havior; however, the noteworthy point is the magni- combined reinforcement provided by response-
tude of their effect on behavior. The other con- dependent reinforcers and attendant stimulus change
spicuous element of these studies is the lack of a that maintains responding for reinforcers in the
common explanation. Conditioned reinforcement presence of equivalent free reinforcers. This stimulus-
most often is adduced to account for stimu1us- reinforcer interaction serves as both a necessary and
reinforcer effects on behavior. Although conditioned a sufficient condition for the emergence and main-You can also read
