Psychol Rec DOI 10.1007/s40732-015-0148-x
ORIGINAL ARTICLE
Blocking in Humans: Logical Reasoning Versus Contingency Learning Diana Delgado 1,2
# Association for Behavior Analysis International 2015
Abstract This study compared the occurrence of the blocking effect when participants had unlimited and limited time to respond to a causal learning task. In contrast to the dominant views in human causal learning, the underlying assumption is that blocking can be sufficiently explained by the same principles that describe conditioning outcomes in animals, but only when logical reasoning about the experimental task is impeded. Experiment 1 compares responses to blocking tests by participants in timed and untimed groups. As expected, most cases of blocking were observed for participants in the timed group. Experiment 2 explores an alternative procedure in which all information about stimulusoutcome associations was simultaneously present. Based on this information, participants sorted target and control stimuli according to their predicted of outcomes. Very limited evidence of blocking was observed with this procedure. Findings are discussed in terms of the interference of rule generation processes with direct contingency control. Keywords Blocking . Contingency learning . Logical reasoning . Causal learning . Pavlovian conditioning . Interference tasks . Rule generation The blocking effect occurs when pairings of a stimulus and an outcome prevent subsequent conditioning of a second stimulus, which is presented simultaneously with the first, and followed by the same outcome (A+, AX+). This procedure * Diana Delgado
[email protected];
[email protected] 1
Fundación Universitaria Konrad Lorenz, Cra 9Bis No 62-43, Bogotá, Colombia
2
Present address: Shellers Bend, # 2, State College, PA 16801, USA
represents a circumstance in which the prediction of an outcome does not depend solely on contiguity relations between stimuli; but is also influenced by previously established competing contingencies (Rescorla & Wagner, 1972). Dominant theoretical models of this phenomenon have accounted for attenuated responses to X when presented alone in terms of its redundancy in the prediction of the outcome (Rescorla & Wagner, 1972; Mackintosh, 1975). Although this effect has been sufficiently demonstrated in animal experiments, results have not been as consistent with humans. The most typical procedure in human blocking research consists of an experimental task in which stimuli (i.e., types of food, drinks, or medicines) are paired with an outcome (e.g., an illness). Once a history of these pairings has been established, a novel stimulus is simultaneously presented with the first stimulus, and both are followed by the occurrence of the illness. Blocking is then evaluated by asking participants to estimate the probability that the second stimulus alone will produce the illness. Because the stimulus–stimulus pairings are easily interpreted in terms of one or two events causing another event (e.g., consumption of food causing an illness), this body of research is commonly described in the literature as human causal learning. Among the studies of cue competition effects in human causal learning, those focusing on the blocking effect seem to have engendered the most notable and impactful amount of empirical research and theoretical controversies (Delgado & Medina, 2013). Two main interpretations have accounted for the variability of research findings regarding the observation of the blocking effect. A dominant interpretation with respect to the high variability of outcomes in human blocking stresses that human causal learning cannot be explained by the same set of contingencies described by Pavlovian animal experiments. This interpretation is based on findings from a series of research studies that
Psychol Rec
report greater blocking when the procedures provide specific information about the predictive function of the target cue X (De Houwer, Beckers, & Vandorpe, 2005; Shanks, 2010; Penn & Povinelli, 2007; Vandorpe & De Houwer, 2006). The blocking effect can be described as an ambiguous circumstance with respect to the causal function of the stimuli in producing the outcome. Thus, from a cognitive standpoint, blocking in humans is more likely when the experimental preparation provides specific information about the relations between each of the stimuli and the outcome. A large body of empirical evidence supports this assertion. Two of the most common procedures used in these studies are (a) those that specify that the stimuli in the compound have additive effects (e.g., Livesey & Boakes, 2004; Lovibond, Been, Mitchell, Bouton, & Frohardt, 2003) and (b) those that specify a submaximal intensity in the occurrence of the outcome (De Houwer, Beckers, & Glautier, 2002; Beckers, De Houwer, Pineño, & Miller, 2005). A study by Livesey and Boakes (2004) illustrates the first case. In Experiment 1, participants were asked to identify which of a variety of chemicals would make a newly discovered bacteria multiply. In the nonadditive group, presentation of both single (P) and compound (PQ) stimuli were followed by a normal effect (+), that is, bacteria increased twice in number as a reaction to the chemical. By contrast, in the additive group, single stimulus presentations were followed by a normal effect (P+), and compound stimulus presentations were followed by a strong effect, that is, the number of bacteria increased four times as a reaction to the chemical (PQ++). As expected, the data showed a significant difference between groups, with higher blocking effects for the additive group. In the second commonly used procedure, the intensity of the outcome is specified. When the magnitude is either unspecified or maximal—that is, if an illness occurs at its maximum possible intensity—the contribution of each stimulus to the outcome is ambiguous. By contrast, when participants are told that the outcome of both single and compound stimuli occur with a submaximal intensity—for example, at 30% of its maximum potential—the redundant role of the target stimulus is stressed and, thus, blocking is likely to occur. For example, Beckers et al. (2005) presented fictitious records of patients who presented an allergic reaction after having eaten different foods. For participants in the submaximal condition the allergic reaction was of moderate strength, while for the other half it was a strong reaction. During blocking tests, participants were asked to rate on a scale from 1 to 9 how likely each of the foods was to produce the allergic reaction. As it has been systematically shown, blocking was stronger for the group in the submaximal condition (Lovibond et al., 2003, Experiment 1; Waldmann & Walker, 2005). Overall, these findings have led authors to conclude that blocking results from inferential processes occurring during
the task rather than from a weakened conditioning of the target stimulus produced by the arranged contingencies (De Houwer & Beckers, 2003; Lovibond et al., 2003). The opposing view argues that human causal learning is sufficiently explained by associative processes or by the principles that describe stimulus–stimulus contingencies. For example, Beckers, Miller, De Houwer, & Urushihara (2006) demonstrated that equivalent results are obtained when manipulations similar to those used in human causal learning tasks are used in studies using animals as subjects. Furthermore, some studies suggest that reasoning about the task may actually hinder rather than facilitate conditioning outcomes. For example, several studies (operant and respondent) have successfully produced conditioning outcomes when contingency awareness is impeded by superimposing a masking task or a cognitive interference task on the presentation of stimulus pairings (see, e.g., Delgado, Medina, & Soto, 2011; Karazinov & Boakes, 2007; Tonneau & González, 2004; Walther & Nagengast, 2006). Restricting time to respond to test trials has also been used to impede reasoning and verbalizations about the objectives of the task. It is assumed that such time restrictions will impede rule-generation processes that may interfere with nonverbal control by the contingencies arranged in the experimental preparation (Karazinov & Boakes, 2007). Accounts of blocking effects based on logical reasoning processes (and not as a result of associative learning processes), seem to vastly outnumber those supporting the contingency learning account, however. In sum, research findings show that greater blocking is predicted when the experimental task provides information that disambiguates the contribution of the target cue on the prediction of the outcome. More specifically, it is argued that manipulations of outcome maximality and additivity show that logical reasoning and rule based processes may be necessary to produce blocking (De Houwer et al., 2005; Livesey & Boakes, 2004; Vandorpe & De Houwer, 2006; Vandorpe, De Houwer, & Beckers, 2005). This interpretation may be questionable, however. Specifying the environmental elements involved in a contingency relation may produce a higher accuracy in predicting the occurrence of blocking effects in both animals and humans. In addition, results from maximality and additivity studies may be interpreted in terms of differential histories of responding to the stimuli in the compound either separately (elemental responding), or as a unique stimulus configuration (configural responding; Glautier, 2002; Livesey & Boakes, 2004; Melchers, Lachnit, Üngör, & Shanks, 2008). Elemental responding is associated with increased blocking effects because it allows for competitor-outcome associations (A+) to weaken or impede subsequent compound-outcome associations (AX+), and not because of the reasoning processes involved.
Psychol Rec
Ascertaining whether or not deliberate reasoning produces different blocking outcomes from those observed in the traditional associative learning literature would entail comparing the results of procedures that facilitate or impede reasoning while maintaining unaltered the predictive ambiguity of the target cue with respect to the outcome. Karazinov and Boakes (2007), for example, included a time restriction for one of two groups exposed to a second order conditioning (SOC) procedure and examined if differential outcomes were observed. The authors found SOC for the group in the time-restricted condition only. Contrary to the mainstream view, these authors discuss that when logical reasoning is obstructed, responding occurs on the basis of associative learning processes. Similar to what has been demonstrated by operant learning experiments on rule-governed behavior (see Catania, 2013), Karazinov and Boakes (2007) suggested that processes of rule generation about the operative contingencies may therefore interfere with direct contingency control. The blocking procedure may be construed as an ambiguous situation with respect to the causal role of the target cue in producing the outcome, wherein deliberate reasoning may act to modulate the occurrence of blocking effects. More specifically, greater blocking should be expected when reasoning is impeded. In agreement with Karazinov and Boakes (2007), the rationale under this assumption is that rule-generation processes may prevent behavioral outcomes from undergoing nonverbal contingency control. In a recent attempt to further examine this hypothesis, Delgado (2013) used two different types of questions to induce causal judgments as a result of intuitive reasoning or logical reasoning (see also García-Retamero & Dieckman, 2006; Todd & Gigerenzer, 2000; Tversky & Kahneman, 1983). These terms are often used in the decision-making literature to describe rule-based (logical reasoning) versus nonrule-based responding in a decision-making situation (Kahneman, 2003; Sloman, 1996). While the question that was oriented toward logical reasoning requested participants to provide a probability rating of the cue producing the outcome, the intuitive reasoning question forced a yes/no answer by asking participants whether the target cue alone would produce the outcome. Although differences in blocking were not found, it is possible that the question types did not facilitate one or the other type of reasoning as intended. Most likely, because logical reasoning was not directly obstructed, participants responded to both questions analyzing the probabilities involved given the situation. The present study attempts to further examine if logical reasoning with respect to the contingencies presented in the experimental task hinders or facilitates the occurrence of the blocking effect. The procedure, which is a variation of the card game simulation used by Glautier (2002), examines differences in blocking in a group with a time restriction to respond
to the target cue, and a group without this restriction. In the first phase of this procedure, a card with background color A (the competitor cue) is followed by a payoff. In the following phase, a card with background color A and figure X (target cue) is also followed by a payoff. In the last phase, a card with figure X and a neutral background color is presented, and participants are required to judge whether or not this card produces a payoff. In his study, Glautier (2002) argued that presenting target and competitor cues as part of one same stimulus object may be associated with a decreased likelihood of blocking, as this stimulus arrangement favors configural responding. However, it has been demonstrated that elemental or configural responding depends on the participants’ history of exposure to one or the other type (Melchers et al., 2005; Melchers et al., 2008). Some aspects of the original task were modified in order to clarify which cues were irrelevant in terms of their outcomes and which cues produced a payoff.
Experiment 1 Method Experimental Design and Data Analysis Most of the experiments conducted in this area evaluate blocking effects by statistically comparing group average probability scores of causal attribution for all the relevant cues. In contrast, we evaluated blocking effects by comparing responses to target, competitor, and control cues from each participant. The traditional blocking procedure including an intrasubject control condition was used. In the first phase of this procedure, an event is paired with an outcome. In the second phase, the same event in combination with a second event is paired with the outcome. A novel compound stimulus followed by the outcome serves as the intrasubject control condition. Data from participants in both groups (timed and untimed) are presented in tables showing individual responses to target, competitor and control stimuli. Participants Participants were 26 undergraduate psychology students. All of them received course credits for volunteering to participate in the experiment, and they were assigned to the untimed (n=13) and timed (n=13) groups on the basis of order of arrival to the laboratory. Informed consent was obtained from all participants at the beginning of the experimental sessions. Data from two participants, who did not complete the experimental task for the timed group, were not included in the analysis.
Psychol Rec
Apparatus and Stimuli The experimental task was presented on IBM-compatible computers with 17-in. screens. All phases of the experiment were designed and implemented using the software LabView, 2010 for Windows. The experiment included 14 cards (7.5×11 cm). Each card included two types of stimuli: a black figure on the center of the card and a background color. Seven background colors and seven figures were combined for a total of 14 cards. Background colors were: A = red, B = green, D = yellow, E = orange, M = gray with horizontal stripes, H = purple, and I = checkered gray. Figures were: N = down arrow, O = clover, P = diamond, Q = heart, X = crescent, F = circle, and R = two-way vertical arrow. Throughout this manuscript, each card will be designated with two letters corresponding to the background color and the center symbol, respectively (e.g., AO). For the cards associated with a payoff, a 7.5×7.5 picture of a bag of money was presented to the right of each card. The same picture, crossed off with a diagonal red line, indicated no payoff. Procedure Participants were informed that the objective of the study was to identify how relations among stimuli may characterize some aspects of human learning. At the start of the experiment, the following instructions were provided on the computer screen: This is a game of cards. The computer will deal one card at a time. As you will see, only some of them produce a payoff. Eventually you may be asked to indicate whether a card produces a payoff or not. Observe carefully. You may now begin. Each training trial consisted on the presentation of a card on the left side of the screen. After 2 seconds, the picture indicating payoff or no payoff appeared 6 cm to the right of the card. Both stimuli remained on the screen for 3 more seconds. The following trial was preceded by a 1.5 seconds intertrial interval. First Training Phase Four cards and their corresponding consequence (payoff/no payoff) were presented four times, in random order: AN+, EN-, MF-, and MX+. This corresponds to the first phase of the blocking procedure in which competitor stimuli are presented separately and followed by the outcome. There are two competitor stimuli in this phase, A and X. In card AN, A is the background competitor stimulus for the figure blocking test, and X in card MX is the figure competitor stimulus for the background blocking test. The function of figure N is rendered null given that it is presented both with the outcome in card AN and without the outcome in card EN.
Similarly, because a figure stimulus cannot be presented in the absence of a background, background color M was used as a neutral cue. It is rendered neutral because it is associated with the prediction of the outcome in MX+ and with its absence in MF-. Once a card completed four presentations, on the fifth trial a test was presented for that card only. In the test trial, two buttons labeled BPayoff^ and BNo Payoff^ were shown next to the card. A correct response was followed by the presentation of the remaining training trials for the other three cards. An incorrect response added three extra training trials for that card and an additional test trial. This correction procedure was repeated until correct outcomes for all cards were selected. No other consequences were provided for correct or incorrect selections.
Second Training Phase The same procedure used in the first training phase was used to train four additional card–outcome associations: AO+, BP+, DQ-, and HX+. This phase corresponds to the second phase of the traditional blocking procedure in which the stimulus presented in the first phase (i.e., competitor stimulus) is now presented simultaneously with a second stimulus (i.e., the blocking or target stimulus), and followed with the outcome (payoff). In this phase, AO and HX are the relevant cards. For the figure blocking test, figure O is now presented with background color A, which reliably predicted the outcome in the first training phase. For the background blocking test, background color H is now presented with figure cue X, which reliably predicted a payoff in the first training phase. Card BP+ serves as a control stimulus because neither of the cues (B or P), were presented previously in association with a payoff; and DQ- is a filler cue (see Table 1). For the timed and untimed groups, these training phases were identical. Thus, to summarize, the figure blocking procedure included AN+ (first training phase) and AO+ trials (second training phase), where O was the figure target cue and A the competitor cue. The background blocking procedure included MX+ (first training phase) and HX+ trials (second training phase), where H was the background target cue and X the competitor.
Table 1
Summary of the design
Experimental Stage
Compound Cues and Outcomes
Training Phase1 Training Phase 2 Tests
AN+, EN-, MF-, MX+ AO+, BP+, DQ-, HX+ MO, HN, BN, MP, IX, AR
Note. The first letter corresponds to the background color of the card and the second corresponds to its figure. O and H are target cues, A and X are their corresponding competitors, and BP is the control stimulus. Payoff is indicated by a (+) sign, and no payoff is indicated by a (-) sign.
Psychol Rec
Tests Prior to the beginning of this phase, the untimed group received the following instruction on the screen: BYou will now see a new set of cards. The computer will deal one card at a time. Please indicate if you think the card pays off or not. Now, let’s play!^ For the timed group, the following additional instruction was included: BYou will have 3 seconds to respond. Please observe each card carefully and try to respond quickly. Now let’s play!^ In this phase target, control, and competitor cues, were tested, by presenting them with neutral background color M or with neutral figure N. The card used to evaluate responding to the figure target cue was MO. Card HN was used to evaluate responding to the background target cue. Their corresponding controls were MP and BN, where M and N were neutral stimuli and P and B were the cues presented as part of the control compound in Training Phase 2 of the blocking procedure. Thus, blocking would occur if (a) a no-payoff response is predicted in the presence of the card with the target cue and (b) a payoff response is predicted in the presence of the card with one of the stimuli of the control compound. Cards IX and AR were used to evaluate predictions with respect to competitor cues X and A. Each card was presented only once to simulate a real card game. On each trial, a card was presented on the left side of the screen. Participants were asked to click on one of three buttons located to the right of the card. Buttons were labeled BPays off, ^ BDoes not payoff,^ or BI don’t know.^ For the timed group, each test trial was available for 3 seconds. If the participant did not respond within this time, the trial was presented again at the end of the randomized sequence. Table 1 shows the experimental phases and the cards presented in each phase. Results In accordance with our hypothesis, blocking effects should occur more consistently in the timed group. That is to say, with the time restriction impeding reasoning about the causal relations and probabilities involved in the prediction of a payoff, responding should come under the direct control of Pavlovian contingencies, and blocking should be reliably observed. Tables 2 and 3 show responses to target (MO, HN) and control (BP) cues for the untimed and timed groups, respectively (background cue M and figure N are neutral cues). Blocking effects were determined on the basis of the caseby-case comparison of responses to target and control cues. More specifically, blocking effects are seen when a nopayoff response is attributed to target cues while a payoff or a Bdon’t know^ response is attributed to control cues. To facilitate visual inspection of the data, blocking effects are indicated by a check mark on the fourth and seventh columns of Tables 2 and 3.
Payoff/no payoff reports for cues O (figure target) and P (control) show differences in blocking effects between the timed and untimed groups for responses to the figure target cue. In the untimed group (n = 11), only one participant showed blocking of the figure target cue, and three participants showed blocking of the target background cue. For the timed group (n=13), blocking of the figure cue was observed for eight participants, and blocking for the background cue was observed for six participants. Four participants (3, 7, 8, and 9 in Table 3) showed blocking for both the figure and the background cue. Tables 4 and 5 summarize these data by showing group differences in blocking effects for figure and background cues, respectively. Chi-square values were calculated using Fisher’s exact test to establish if observed differences between groups were statistically significant. As it is shown in Table 4, blocking of the figure cue was observed for a higher number of participants in the timed group (8/13) as compared with the untimed group (1/11). This difference was statistically significant, with χ2 =6.99, p=.01. A similar finding was observed for the background cue. Although this difference was not statistically significant χ2 =0.90, p=.42, more cases of blocking of the background target cue were observed in the timed group (6/13) than in the untimed group (3/11). By using a correction procedure during Training Phase 1 and 2, we guaranteed that responses to the target stimulus occurred after participants had successfully acquired competitor–outcome associations. In addition to testing responses to the target stimuli we also examined the extent to which the associative strength of the competitors acquired during phase 1 (A+ and X+) and phase 2 (AO+ and HX+) would generalize to the novel compounds AR and IX in a test trial. For the untimed group, 9 of 11 participants reported that they were unsure as to whether or not the novel card AR produced a payoff, one participant responded that it produced a payoff, and one participant responded that it did not. With respect to the IX compound, six participants responded that it produced a payoff and five responded that they were unsure. For the timed group, 9 of 13 participants responded they were unsure about the outcome of card AR, and four responded that the card did not produce a payoff. For the card IX, seven participants responded that it produced a payoff and six participants were unsure about the outcome. Discussion The data presented in Tables 2 and 3 show greater blocking in the timed group. Approximately half of the participants of this group showed blocking of the target cues. By contrast, only one participant in the untimed group showed blocking for the target cue O, and three participants showed blocking for target cue H. These differences suggest that, as expected, when deliberate reasoning about the
Psychol Rec Table 2 Blocking/no blocking effects of figure and background target cues in the untimed group
Ppts
O (figure target cue)
P (control figure)
Blocking of figure cue
H (background target cue)
B (control background)
Blocking of background cue
1 2
− +
+ +
✓
+ −
+ +
✓
3
+
d/k
−
d/k
✓
4 5
+ +
+ d/k
+ +
+ +
6 7
+ +
d/k +
+ +
d/k +
8
+
+
+
d/k
9 10
+ +
d/k +
+ +
+ +
11
+
+
−
d/k
✓
Note. Payoff is indicated by a (+) sign, and no payoff is indicated by a (-) sign. The letters d/k indicate that participants were unsure about the outcome of the card.
task is impeded, responding may be more likely to resemble the patterns observed in animal experiments, that is to say, behavior may be more likely to fall under the control of the principles of contingency learning. However, the fact that half of the participants in the timed group did not show blocking shows that additional factors are involved in the prediction of responding with respect to compound stimuli. Configural and elemental histories of responding may constitute one such influential factor. On the other hand, the lack of negative predictions by control cards in the figure blocking test (participants in both groups responded that the control card produced a payoff and a few of them gave a Bdon’t know^ response) shows that previous exposure to only one of the elements of the compound (in Training Phase 1) does affect causal judgments of Table 3 Blocking/no blocking effects of figure and background target cues in the timed group
the cues in the compound. This effect may be explained in terms of a transfer of associative strength between the previously exposed cue (competitor), the added element (target), and the outcome (Rescorla & Colwill, 1983). In both groups, the number of Bdon’t know^ or uncommitted responses occurred almost exclusively for control cues. While these responses may constitute evidence of elemental responding and cognitive reasoning, they may also be indicative of weak associative strength acquired by the cues during the sequence of training trials. To examine the latter possibility we contrasted causal judgments of target, competitor, and control cues, using an alternative procedure. In Experiment 2 all the information regarding cues and outcomes was simultaneously present on the screen. Cards from Training Phases 1 and 2 were presented in a payoff and no payoff box,
Ppts
O (figure target cue)
P (control figure)
Blocking of figure cue
H (background target cue)
B (control background)
1 2 3 4 5 6 7 8 9 10 11 12 13
− − − + + + − − − + − + −
+ d/k + d/k + d/k d/k d/k + + + + d/k
✓ ✓ ✓
+ d/k − d/k + − − − − − + + +
+ d/k + d/k + d/k + d/k d/k d/k d/k + +
✓ ✓ ✓ ✓ ✓
Blocking of background cue
✓
✓ ✓ ✓ ✓ ✓
Note. Payoff is indicated by a (+) sign, and no payoff is indicated by a (-) sign. The letters d/k indicate that participants were unsure about the outcome of the card.
Psychol Rec Table 4 Frequency distribution of blocking effects for the figure cue in the timed and untimed group Group
Blocking of figure cue
No blocking of figure cue
Total
Untimed
1
10
11
Timed
8
5
13
Total
9
15
24
χ =6.99, p=.01.
were designed and implemented using the software LabView, 2010 for Windows. The experiment included the same cards described in Experiment 1. The size of the cards was adjusted (2.5×3.5 cm) for this procedure so that all the cards could fit into one screen. Procedure
2
respectively. Based on this classification, participants were asked to place the unclassified test cards into one of the boxes.
Experiment 2 One of the most widely used procedures to study the acquisition of contingency relations in humans is the discrete trial procedure. Even though measures are usually taken to ensure that the relevant relations are sufficiently trained, in human experiments there is a chance that factors such as attention, motivation, or habituation to the conditions of the task interfere with the acquisition of a novel repertoire of arbitrary relations. By testing an alternative procedure we examined if differences in blocking were observed when competitor–outcome and target–outcome associations were simultaneously available than when presented across a limited sequence of discrete trials. Method Participants Ten undergraduate students between ages 18 and 30 years volunteered to participate in this study. All participants received bonus points in one of their course assignments upon completion of the experimental task. Apparatus and Stimuli The experimental task was presented on IBM-compatible computers with 17-in. screens. All phases of the experiment Table 5 Frequency distribution of blocking effects for the background cue in the timed and untimed group Group
Blocking of background cue
No blocking of background cue
Total
Untimed Timed Total
3 6 9
8 7 15
11 13 24
χ2 =0.90, p=.42.
As in Experiment 1, participants were informed that the study examined some of the circumstances under which human learning occurs. The following instructions were provided on the computer screen: This is a game of cards. You will notice that some cards produce a payoff and some cards don’t. Observe carefully. On the next screen you will see three boxes labeled BPayoff,^ BNo Payoff,^ and BI Don’t Know.^ Your job is to classify the six cards that appear on the top right corner of the screen into one of the three boxes. You’ll have to try to identify if the cards produce a pay off or not. Using the mouse, drag each card into one of the boxes. When you are ready to begin, click on the button labeled BStart^ at the bottom of the screen. There was no time limit to complete the sorting task. Three boxes were presented on the computer screen. The box labeled BPayoff^ contained the cards AN, AO, BP, MX, and HX. These were the same cards shown as producing a payoff during Training Phases 1 and 2 in Experiment 1. The cards inside the BNo Payoff^ box were EN, DQ, and MF. These were the same cards that resulted in no payoff in Experiment 1 (see Table 1). The cards classified in these two boxes could not be moved. The cards used for the test phase in Experiment 1 (MO, HN, BN, MP, IX, AR) were presented outside the boxes and on the top right side of the screen. These cards could be placed anywhere on the screen using the computer mouse. A button labeled BDone^ on the top center of the screen ended the experiment once the participant had reached a final classification of all test cards. Results Table 6 shows classifications of target and control cues according to the attributed outcome of the cards. Nine of 10 participants classified the MO (target) and the MP (control) card in the Payoff box. Hence, as in the untimed group (Experiment 1), blocking effects were not observed for the figure target cue in any except one participant. By contrast, only two participants classified the background target cue (H) in the Payoff box. One participant classified it in the Don’t Know box and the seven remaining participants classified it in the No Payoff box. Differences in responding to target figure and target background cues were statistically significant
Psychol Rec Table 6 Blocking effect of target and background cues after exposure to the sorting task
Ppts
O (figure target cue)
P (control figure)
1 2
+ +
3
Blocking of figure cue
H (background target cue)
B (control background)
+ +
− −
− −
+
+
d/k
−
4 5
+ +
+ +
− −
+ −
✓
6 7
+ +
+ +
− +
d/k −
✓
8
+
+
+
+
9 10
+ −
+ d/k
− −
− +
✓
Blocking of background cue
✓
Note. Classification in the Payoff box is indicated by a (+) sign, and classification in the No Payoff box is indicated by a (-) sign. The letters d/k indicate that participants classified the card in the Don’t Know box.
(McNemar p=.01). However, because most participants also attributed a negative outcome to the control background cue (B) and a positive outcome to the control figure cue, blocking effects were confirmed only in 3 of 10 participants (see Table 6). No significant differences were found between blocking of target and background cues (McNemar p=.50). Overall, most participants attributed a positive outcome to figure cues (target and control), and a negative outcome to background cues (target and control). In this procedure, competitor–outcome associations AN+, AO+ (for background cues) and MX+, HX+ (for target competitor cues) were apparent as these cards were presented inside the Payoff box along with their corresponding target cue cards (MO and HN). As in Experiment 1, Experiment 2 examined if participants attributed a positive outcome to novel compounds AR and IX, which included the competitor cues A and X. Seven participants classified the novel card AR (A being the background competitor of the figure target cue) in the No Payoff box, and three classified it in the Don’t Know box. By contrast, all participants classified the IX card (figure competitor of background target cue) in the Payoff box. Discussion This procedure produced very weak evidence of blocking. There was a notable difference however between responses to figure and background cues. In general, while all figure cues (target, controls, and competitors) seemed to predict a payoff, background cues tended to predict no payoff. Hence, with this procedure, participants seemed to produce a judgment of outcome based on the figure cue and not on the background cue. Because figure cues were perceptually more salient than background cues, participants may have generalized the information from the cards where the figure cue produced an outcome and derived a rule that was applicable to all cards.
According to this interpretation, responding when there is time to think (even Don’t Know or uncommitted responses) can be explained in terms of the evaluation of the logical possibilities implied in the contingencies, rather than in terms of the strength of the associations.
General Discussion The purpose of this study was to examine if blocking is attenuated by deliberate reasoning in a causal learning task. In order to do this, causal judgments of target and control cues in participants who had three seconds to respond to the blocking test, and in participants who had unlimited time to respond, were compared. Results show notable differences between the two groups. In contrast to Glautier’s (2002) findings, nearly no cases of blocking for both target cues (figure and background) in the untimed group were observed. Most participants in this group associated both cues as well as the control stimuli with a payoff. In contrast, blocking was observed for two thirds of the participants in the timed group. These results suggest that human behavior can be described by the same set of conditioning principles that govern animal behavior, but only when logical reasoning is impeded or highly limited. Most uncommitted or Bdon’t know^ responses correspond to predictions of the outcome of control cues, and the majority of these responses were given by participants in the timed group. This may indicate that participants responded separately to the stimuli in the compound (as they did for the competitor-target compound), even though they were presented as parts of a single stimulus object. In contrast to target and competitor cues, neither of the cues in the control card had been previously associated with the outcome. As a result, participants did not have enough information on which to base a causal judgment with respect to the elements of the control
Psychol Rec
compound card. The same occurred for the evaluation of competitor cues with a novel figure or background. A high percentage of Bdon’t know^ responses further suggests elemental responding with respect to each of the stimulus in the card. These findings show that despite the fact that target and competitor cues were presented as part of a single stimulus, participants responded elementally with respect to each of the cues in the compound. In contrast to most views, though, we suggest that responding either configurally or elementally does not necessarily facilitate blocking. What does seem to account for blocking effects in experimental preparations designed to produce elemental responding (i.e., Beckers et al., 2005; De Houwer et al., 2005), is that in these procedures, the arranged contingencies are such that the null contribution of the target cue is specified. In order to rule out if Bdon’t know^ responses were due to insufficient exposure to training trials, we examined responses to the test cards using a sorting-task procedure. By contrast to the traditionally used procedures, in this task, information regarding the outcomes of the cards was not presented on a trialby-trial basis. In this procedure, the cards were classified into two boxes according to their outcome. Participants in this group responded differently than those in the former two groups (Experiment 1). Very few cases of blocking were observed for both target cues (figure and background). Most participants attributed payoff outcomes to figure cues, and no payoff outcomes to background cues. Additionally, in contrast with the timed and untimed groups, very few cases of uncommitted responses to the cards were observed. Most likely, this experimental arrangement is associated with a higher likelihood of self-generated rules with respect to the task. Seemingly, these rules produce behavioral outcomes different from those produced by the temporal associations involved in discrete trial contingencies. It would be interesting to examine if limiting the time to sort the test cards would result in a different pattern of responding as in Experiment 1. Further research could focus on the comparison between the sorting task and the discrete trial procedure. On the other hand, it is possible that the inclusion of two blocking tests (one for a figure cue and one for a background cue) may have increased the overall difficulty of the task. Future replications of present findings should be conducted using one blocking procedure only. This study provides empirical evidence contributing to the discussion of whether or not the principles of classical conditioning that have been established in the animal laboratory, the blocking effect in particular, are also descriptive of human behavior under similar circumstances of contingency control. The effect of verbal processes on nonverbal contingency control procedures is without doubt crucial in determining the generality of behavioral processes and outcomes and, in particular, in furthering our understanding of human behavior in light of such processes.
To summarize, results showed greater evidence of blocking for participants exposed to the procedure in which cognitive reasoning about the task was impeded. This can be explained in terms of the operation of rule-based processes, which may in some cases compete with the control exerted by the contingencies arranged in the experimental task. Thus, when logical reasoning is taken out of the equation, principles of contingency learning seem sufficient to describe human behavior in causal learning tasks.
References Beckers, T., De Houwer, J., Pineño, O., & Miller, R. R. (2005). Outcome additivity and outcome maximality influence cue competition in human causal learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31, 238–249. doi:10.1037/ 0278-7393.31.2.238. Beckers, T., Miller, R. R., De Houwer, J., & Urushihara, K. (2006). Reasoning rats: Forward blocking in Pavlovian animal conditioning is sensitive to constraints of causal inference. Journal of Experimental Psychology: General, 135, 92–102. doi:10.1037/0096-3445.135.1.92. Catania, C. (2013). Learning (5th ed.). Cornwall-on-Hudson, NY: Sloan. De Houwer, J., & Beckers, T. (2003). Secondary task difficulty modulates forward blocking in human contingency learning. Quarterly Journal of Experimental Psychology, 56B, 345–357. De Houwer, J., Beckers, T., & Glautier, S. (2002). Outcome and cue properties modulate blocking. Quarterly Journal of Experimental Psychology, 55A, 965–985. De Houwer, J., Beckers, T., & Vandorpe, S. (2005). Evidence for the role of higher-order reasoning processes in cue competition and other learning phenomena. Learning & Behavior, 33, 239–249. Delgado, D. (2013). Magnitude and valence of outcomes as determinants of causal judgments. Revista Latinoamericana de Psicología, 45(1), 9–20. doi:10.14349/rlp.v45i1.1307. Delgado, D., & Medina, I. F. (2013). Cuando la contigüidad no es suficiente: Bloqueo en relaciones de equivalencia. Universitas Psychologica, 12(2), 613–626. Delgado, D., Medina, I. F., & Soto, J. S. (2011). El lenguaje como mediador de la transferencia de funciones: Es necesario nominar para relacionar? Revista Mexicana de Análisis de la Conducta, 37, 31–52. doi:10.5514/rmac.v37.i2.26138. García-Retamero, R., & Dieckmann, A. (2006). Una visión crítica del enfoque de los heurísticos rápidos y frugales. Revista Latinoamericana de Psicología, 38(3), 509–522. Glautier, S. (2002). Spatial separation of target and competitor cues enhances blocking of human causality judgments. The Quarterly Journal of Experimental Psychology, 55B, 121–135. Kahneman, D. (2003). A perspective on judgment and choice: Mapping bounded rationality. American Psychologist, 58, 697–720. Karazinov, D. M., & Boakes, R. A. (2007). Second order conditioning in human predictive judgments when there is little time to think. Quarterly Journal of Experimental Psychology, 60, 448–460. doi: 10.1080/17470210601002488. Livesey, E. J., & Boakes, R. A. (2004). Outcome additivity, elemental processing and blocking in human causality judgments. The Quarterly Journal of Experimental Psychology, 57B(4), 361–379. Lovibond, P. F., Been, S. L., Mitchell, C. J., Bouton, M. E., & Frohardt, R. (2003). Forward and backward blocking of causal judgment is enhanced by additivity of effect magnitude. Memory & Cognition, 31, 133–142. doi:10.3758/BF03196088.
Psychol Rec Mackintosh, N. J. (1975). A theory of attention: Variations in the associability of stimuli with reinforcement. Psychological Review, 82, 276–298. Melchers, K. G., Lachnit, H., Üngör, M., & Shanks, D. R. (2005). Prior experience can influence whether the whole is different from the sum of its parts. Learning and Motivation, 36, 20–41. doi:10. 1016/j.lmot.2004.06.002. Melchers, K. G., Shanks, D. R., & Lachnit, H. (2008). Stimulus coding in human associative learning: Flexible representations of parts and wholes. Behavioural Processes, 77, 413–427. Penn, D. C., & Povinelli, D. J. (2007). Causal cognition in human and non-human animals: A comparative, critical review. Annual Review of Psychology, 58, 97–118. doi:10.1146/annurev.pch.58.110405. 085555. Rescorla, R. A., & Colwill, R. M. (1983). Within compound associations in unblocking. Journal of Experimental Psychology: Animal Behavior Processes, 9(4), 390–400. doi:10.1037/0097-7403.9.4. 390. Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlovian conditioning: variations in the effectiveness of reinforcement and nonreinforcement. In A. H. Black & W. F. Prokasy (Eds.), Classical conditioning II: current research and theory (pp. 64– 99). New York, NY: Appleton-Century-Crofts. Shanks, D. R. (2010). Learning: From association to cognition. Annual Review of Psychology, 61, 273–301. doi:10.1146/annurev.psych. 093008.100519.
Sloman, S. A. (1996). The empirical case for two systems of reasoning. Psychological Bulletin, 119, 3–22. Todd, P. M., & Gigerenzer, G. (2000). Précis of simple heuristics that make us smart. Behavioral and Brain Sciences, 23, 727–780. Tonneau, F., & González, C. (2004). Function transfer in human operant experiments: The role of stimulus pairings. Journal of the Experimental Analysis of Behavior, 81, 239–255. doi:10.1901/ jeab.2006.49-05. Tversky, A., & Kahneman, D. (1983). Extensional versus intuitive reasoning: The conjunction fallacy in probability judgment. Psychological Review, 90(4), 293–315. doi:10.1037/0033-295X. 90.4.293. Vandorpe, S., & De Houwer, J. (2006). People want to see information that will help them make valid inferences in human causal learning. Memory & Cognition, 34, 1133–1139. doi:10.3758/BF03193259. Vandorpe, S., De Houwer, J., & Beckers, T. (2005). Further evidence for the role of inferential reasoning in forward blocking. Memory & Cognition, 33, 1047–1056. Waldmann, M. R., & Walker, J. M. (2005). Competence and performance in causal learning. Learning & Behavior, 33, 211–229. Walther, E., & Nagengast, B. (2006). Evaluative conditioning and the awareness issue: Assessing contingency awareness with the fourpicture recognition test. Journal of Experimental Psychology: Animal Behavior Processes, 32, 454–459.