11/2/99
Joint control and rule following:  An analysis of purpose
Barry Lowenkron
Presented at  the annual convention of the Association for Behavior Analysis, Chicago, May, 1999.
 Return to Table of Contents

It has often been observed that humans encounter many novel situations having already been instructed regarding the contingencies that operate in these situations.  In such cases, prior to the first reinforcer or the first punisher,  these instructions play a unique role in controlling behavior.  It is a role that cannot be repeated once the behavior has been emitted and exposed to an effective consequence, for now this consequence also exerts control over the behavior.  Thus, it is in the initial exposure to an instructed situation that the instructions play a unique role, and it is to the explication of behavior in this situation, the situation that exists after instructions have been given, but before any effects of reinforcement or punishment are manifest, that this paper is directed.

 The nature of these instructions have been variously characterized as rules or as contingency-specifying stimuli.  Thus, for Skinner, a rule is a verbal statement specifying all or part of an actual contingency of reinforcement.  The rule is seen to exert control over behavior by acting as an SD  (Skinner, 1969).  It is often recognized however, that there are problems with this formulation, for a  statement may effect behavior without the history of differentially reinforced responding that defines SD control; as for example where a person, making a cake for the very first time, conforms to the recipe instruction  “When the cake rises, remove it from the oven.”

Furthermore, as Schlinger and Blakely have observed, rules often function in ways that SDs do not, as for example where they alter the functions of subsequent stimuli.  Thus the statement “When the cake rises, remove it from the oven” alters the function of the cake; conferring upon the risen cake the power to evoke the behavior of removing it from the oven.

But even with this function-altering formulation, there is still the question of just how rules cause certain subsequent stimuli (but not others) to control behavior. This is the question addressed here:  How do rules control behavior?    That is, how do statements heard at one time, describing contingencies of reinforcement, then control later behavior?  To begin an answer, let us first identify what seem to be the main impediments to a behavioral account of rule-governed behavior and then explore some answers.  Though for the sake of clarity and brevity the terminology describing these impediments may be cognitive, it is intended that the answers  be rigorously behavioral.

FIG.1 Three problems of a rule:  Memory, recognition  and response

Let us continue to use as our illustration the rule “When the cake rises, remove it from the oven.”  As shown at the top of Figure 1, the rule statement is heard at one time, and then, at some later time, the event specified by the rule actually occurs, and the cake rises.  This event, the risen cake, then evokes  the specified response; here, removing the cake from the oven.

Now the first problem for a behavioral account is in explaining what we may call the memory function: How is the effect of hearing the rule statement preserved over time?  That is, how does hearing the rule statement at one time affect behavior even hours after the statement was heard?

The second problem has to do with the recognition function: How is the event specified by the rule actually recognized when it occurs?  For example, when the cake finally rises, how is this event recognized as the particular event mentioned in the statement of the rule that was heard, perhaps, hours before?

Thirdly, there is the problem of the response function.  Thus, assuming the subject recognizes the specified event, here the cake rising,  how is it that the specified event, and no other, then evokes the response specified in the rule, so that only when the cake rises does the subject make the specified response and  remove the cake from the oven?

 To begin to provide these answers, however, we must first elaborate on the notion of joint control, for this notion is fundamental to the account presented here.  Joint control in fact involves nothing more than the usual kind of operant stimulus control, except that under joint control two verbal stimuli exert stimulus control over a, single, common verbal topography.   Now while this may not sound like much, this form of stimulus control seems to have exceedingly powerful and important effects.

(FIG2.) Number array

 Consider, for example, the following task in which you must locate the correct set of numbers in the array shown here in response to the spoken sample 939173.  Take a moment to find the set 939173 in this array;  finding the correct set required joint control. This next figure illustrates what this is.
 
 

FIG 3:   Verbal behavior relations in joint control

First off, notice that as you perused the array, you occasionally rehearsed the sample series 939173 vocally .  In the language of Skinner’s verbal operants, this behavior would be described  as echoic rehearsal of the sample.  Second, as you perused the array, you also attempted to tact the sets of numbers in the array by emitting the echoic 939173, and you continued to do this until, as illustrated here, you encountered a set of numbers that you could emit as a series of tacts from the array, while you simultaneously emitted the same series of topographies as the rehearsed self-echoic.

That is, at some point you could say 939173 both from memory and jointly, as you read those same numbers from the screen.  And so at this point you were repeating the number-series under joint tact/echoic control. This event, this onset of joint control, is a unique source of stimulus control because it only happens when the specified comparison is actually encountered.  Indeed, the onset of this joint self-echoic/tact control over the rehearsed  topography 939173 was in fact the only possible way you could identify which number set in the array was the one initially specified.

It is also important to note here that even although this was a matching to sample task, you were able to recognize the specified number set without actually emitting a pointing response to that number. That is, unlike a pigeon, you did not rely on some SD  to evoke a selection response that in turn disrupted the searching behavior.  Rather, you simply relied upon the onset of joint control to determine when you had located the number set - an event you could report by saying "I found it".  Indeed, the onset of joint control constituted the event you would identify as the event of actually recognizing the specified number set.

But, as illustrated in the lower part of the figure, had there been a request that you point to the correct number set,  you could have easily conformed to this mand, and pointed to that set.  As discussed elsewhere, (Lowenkron, 1998) such a pointing  response may be described as a selection-based  autoclitic response, a response  reporting to others which number in the array entered into joint stimulus control with the self-echoic you were rehearsing.

As we shall soon see, however, more relevant here is the fact that pointing in response to such a mand may also serve to avoid the aversive consequences generally associated with failure to conform to the mands of others.

FIG 4 Another example: Pentagons

  On this next figure, the sample is the phrase black dot in a smaller pentagon.  Again, the comparisons are perused as the sample phrase is rehearsed as a self-echoic.  When a comparison is found that allows the same topography,  black dot in a smaller pentagon to be emitted both as a tact of the comparison,  and also as an echoic of the sample, the source of this joint echoic/tact control, the appropriate comparison,  is reported by an autoclitic pointing response.  Together, this example and the prior number-finding task illustrate the generic nature of joint control.  That is, the joint control event is independent of the properties of any particular stimuli: in these two cases it occurs when a tact and an echoic come to control the emission of a common topography regardless of the particulars of the topography.

(Fig 5)  Application to the  three functions.

 Let us now apply the foregoing.  How does the notion of joint control apply to the memory, recognition, and response problems that we have identified in explaining how rules govern behavior?  Consider the rule told  to a novice baker: When the cake has risen, remove it from the oven

The memory function.  As you see in Figure 5, after the rule statement is heard, it is susceptible to echoic and/or self-echoic control: prior repetitions of the rule control subsequent repetitions. As I have discussed elsewhere, rehearsala variety of different aspects of the environment may control the rate of these rehearsals,  and so, in the present case, let us appreciate the fact that the subject is following the rule because he has never baked a cake before. And so, self-echoic rehearsals of the instructions occur at least each time the subject performs the unfamiliar task of checking the cake to see if it has risen while saying something like, “OK, has it risen?”  This suggests that the memory function, the preservation of the effectiveness of the rule statement to a later time, may be simply described as intermittent self-echoic rehearsal. Though this may seem to be an simple answer,  it has an important benefit: describing self-echoic rehearsal as the solution  to the memory question,  provides for an easy answer to the second question, the nature of the recognition function.

The recognition function  The question here is how a subject recognizes, at some later time, the particular event mentioned at some earlier time by the rule.
And so, as I have illustrated here, eventually the cake rises.  How then is this event, (and no other--a separate problem), recognized as the particular event mentioned earlier in the rule?   I would suggest the following:  When the cake rises, it changes the nature of the stimulus control exerted over the phrase  “cake has risen”.  For now that phrase  may be emitted both as  the self-echoic I just mentioned, and also  as a tact of the now-risen cake.  As a result, the phrase “cake has risen” is emitted under joint tact and self-echoic control.  And just as in the prior examples with the numbers and with the dot in the smaller pentagon, here too, as the subject emits the phrase “cake has risen” while emitting the same topography as a self-echoic, the event of the rising cake may be said to have been recognized by the subject as the event mentioned by the rule.

The response function. Now, having recognized a particular event as the one specified in the rule, the subject must respond, and do what the rule says. This brings us to the third problem: the response function.  How is the particular response mentioned in the rule actually evoked?  To understand this one must first recognize that mention of the response in the rule functions as a mand.  That is, the phrase “remove the cake from the oven”, is a mand.  And, as discussed earlier, once manded, the listener does not remain actionless.  For doing so, and not fulfilling the mand, generally results in punishment, however mild, by the speaker.  The mand is thus a reflexive CEO: the listener must act to terminate the relevant features of the current stimulus situation, and the contents of the mand specify how to do this.

Now in the rule considered here, the phrase “remove the cake” is a mand, but we see that this mand  is not to be acted upon when the rule is first heard, but only after the subject recognizes the fact that the cake has risen. The question then is how does a subject’s recognition that the cake has risen then evoke the manded behavior of removing the cake from the oven?

Now we have already identified the subject’s recognition of the cake rising with the onset of joint control itself,  and so perhaps the onset of joint control  acts as a go/nogo conditional stimulus: one correlated with the opportunity for the subject to successfully fulfill the contents of the mand.  Thus, when the cake is recognized as having risen, this is the moment to obey the mand and remove it from the oven. Doing so earlier would not have been reinforced, and not doing so now leaves the threat of an aversive stimulus for not obeying the rule.  Only upon the occurance of joint control is the specification of the response  “remove from oven” an SD for emitting that behavior.

In summary, to this point it seems that an account phrased solely in terms of verbal operants can describe how the effect of a rule can be maintained over time, how the events it mentions can be recognized, and how they may subsequently be acted upon.

FIG 6  Two kinds of verbalized contingencies

There is  however, a second, and related, kind of rule: one with a verbalized consequence. Thus, as we see at the top here, with the actual cake available, the subject is told, “Bake the cake until it rises”  In this case the antecedent is an actual event or situation, (here the actual cake) to which the subject is given a verbalized response (bake the cake), and that is followed by a verbalized consequence  (till it has risen). For the sake of comparison below this I have put the kind of rule we first discussed (i.e., “When the cake has risen, remove it from the oven”), and we see that in this case it is the antecedent that is verbalized, and the consequence that is actual; Thus we see here two ways in which the three term contingency may be turned into a rule: one in which the antecedent is actual, and the consequence is verbalized, and one in which the consequence is actual, but the antecedent is verbalized -- in both cases the response is  verbalized.

FIG 7 The Verbalized  Consequence

As we see in this next figure, as in the case with the verbalized antecedent, here too, with the verbalized consequence, the components of joint control help mediate the three rule functions: memory, recognition and response.

The memory function.  At the top of this figure we see that hearing the statement “bake until the cake has risen” provides the basis for subsequent self-echoic rehearsal of the verbalized consequence  “cake has risen”. And so, as having heard the statement, our novice baker proceeds with the unfamiliar task of checking the cake occasionally for the specified rising;  perhaps by asking himself if the cake has risen.  The subject continues to check the cake until the phrase “cake has risen” can be emitted as a tact, thereby allowing the term ”cake has risen” to be emitted under joint tact/ self-echoic control.

The recognition function. And as with the prior rule, here too,  I would identify emission of the term “cake risen” under joint control as the moment the subject may be said to have recognized that the event specified in the rule, the cake rising, had actually happened.  And again, as in the prior case, here too, the onset of joint control serves to control subsequent behavior: in this case, to terminate baking.

Because this paper does not focus on the function of the joint control event as a conditioned reinforcer,  it must suffice here to merely point out that under the current circumstances joint control occurs just prior to reinforcement. Thus, it is upon recognizing that the cake has risen that the subject has fulfilled the mand to bake, and avoided threats inherent in not fulfilling the mand.  Under these circumstances it is certainly reasonable to suspect that the onset of joint control functions as a conditioned negative reinforcer.

 FIG 8   Concatenation:

 The foregoing really  provides the beginning rather than the end to this kind of interpretion of rule-governed behavior, and so to close let's see what the next step in this interpretion might look like, for something very interesting happens when a verbalized antecedent and a verbalized consequence are concatenated.  Thus, as we see in this next slide, the subject is given the verbalized consequence “bake until the cake has risen”  followed by the verbalized antecedent  ”when the cake has risen, remove it from oven.”

As the slide shows, the result of this concatenation is to make the joint control event serve as an intermediate link in a chain.  Thus, recognition of the cake’s rising is the event that terminates the first link, baking; and simultaneously sets off the second link: removing the cake from the oven.  Joint control can thus act to produce a new kind of chained behavior:  instead of the the usual kind of behavioral chain, where it is events in the external environment that serve as reinforcers for prior responses and  as SDs for subsequent responses, in the case shown here it is not the event itself, but rather the onset of joint control, that serves as the reinforcer for the prior behavior, and as the initiator of the next behavior.

And though one may not claim this demonstration is exhaustive, it is still what seems to be a simulation of purpose: for with no prompting beyond the initial statement of the rule, the subject behaves so as to produce the event specified by the rule, and upon doing so, recognizes the specified event as having occurred, and then changes behavior so as to move on to the next goal.   In behaving under the control of this concatenation of  rules, the subject might be said  to be behaving with purpose.