Relational Frame Theory 1

Relational Frame Theory

Over the best part of the last three decades, an increasing number of behaviour analysts
have been developing experimental preparations that generate complex and "novel"
human behaviour under laboratory conditions. This research has opened up exciting new
vistas of research for the behavioural analysis of human language and cognition. We will
now outline these research developments.

Many readers will be aware of the well established behavioural principle of Pavlovian or
respondent conditioning. A dog, for example, will become excited when it hears the sound
of its owner's car engine, because on previous occasions hearing this sound has been
followed by the actual arrival of the owner. The same dog may show fear, however, if the
owner shouts in an angry tone, because such shouting has previously been followed by
punishment. Furthermore, we can train a dog to get excited when it hears a specific word,
such as "cookie", by consistently giving the dog some food after saying "cookie". In this
way we can attach important psychological functions (e.g., of food) to previously neutral
events (e.g., "cookie").

Something interesting happens when we reverse this order of events, however. Imagine
that every time we feed a dog with a biscuit we say, "cookie" just after he has finished
eating. When we have done this several times, will the dog become excited (anticipating a
meal) if we say "cookie" without showing him a biscuit? The answer is no. A large body
of research has shown that animals do not readily learn about neutral events, such as
words, that follow important ones such as food (see Hall, 1996). Animals can only easily
learn about events that predict the onset of something which is psychologically important.

For verbally-able humans it is a quite different story. Imagine, for instance, that we
repeated the above experiment in the following way with a young child. Each time we
give the child a cookie we say "cookie" just after the child finishes eating. What would
happen if one day we shouted "cookie" when the child was in a nearby room? Most likely,
the child would come running to us expecting to get a cookie. In effect, the sound of the
word would make the child think of cookies, even though the word "cookie" had never
predicted the delivery of an actual cookie. This is entirely consistent with a large body of
experimental evidence that has shown that humans, unlike animals, have a strong
tendency to relate a neutral event to an important event, even though the former has
always followed the latter. Respondent conditioning, therefore, is often radically different
for verbally-able humans than for all other animals. When the word cookie predicts the
delivery of an actual cookie, both humans and non-humans can quickly learn to become
excited. Only for the human, however, does the word cookie and the actual cookie enter
into a bi-directional stimulus relation wherein each can equally stand for the other. For the
new-wave behaviour analyst, this bi-directionality is deemed to be one of the most
important defining features of human language and cognition.

Another important feature of human language and cognition, from the new-wave
perspective, involves the emergence of complex networks of related events. Imagine, for
example, a young girl who eats a cookie. Afterwards she is told, "You have just eaten a
cookie, and another word for cookie is biscuit". From now on, whenever she hears the
word "biscuit" she will probably think of the word "cookie," and actual cookies as well. In
effect, simply hearing the word "biscuit" can make the girl think of an actual cookie, even
though the word has never been directly associated with a real cookie. Numerous studies
 Relational Frame Theory 2

have demonstrated this basic effect, and have also shown that it is possible to teach even
young children large and complex relational networks (e.g., Smeets, Barnes & Roche,
1997). When this occurs, we say that an equivalence relation has been established
between actual cookies, the word "cookie" and the word "biscuit".

The construction of relational networks, such as equivalence relations, between words and
events seems to underlie many facets of human language and cognition. Mathematics, for
example, is the result of thousands of years of developing and refining increasingly
complex and abstract relational networks. The logical statement "If A = B and B = C then
A = C" represents just one very simple relational network that tells me the value for C
based on the value for A (i.e., A and C participate in a derived transitive relation). With
this simple network, if I weigh A and find it be 1 kg, I now know that both B and C each
weigh 1 kg without having to weigh them.

Relational networks are also exciting because they appear to parallel many natural
language phenomena, including, for example, naming behaviors. For instance, if a young
child is taught to point to an object given a specific written word, the child may point to
the word given the object without further training. Consequently, given training in the
spoken word "chocolate" and actual chocolate, and between the written word chocolate
and the spoken word "chocolate," a child will identify the written word chocolate as in an
equivalence relation with "chocolate," even though this performance has never actually
been trained. Thus, symmetry and transitivity between written words, spoken words,
pictures and objects is commonplace in naming activity (Hayes, Gifford & Ruckstuhl,
1996).

How is relational responding established? We should be clear at this point that the
description of language and cognition in terms of relational networks, does not, on its
own, constitute a behaviour analytic explanation of these important human phenomena. In
order to explain language and cognition (e.g., derived relations between written and
spoken words) we use RFT (Hayes & Hayes, 1989) which seeks to explain the generative
nature of language in terms of already-established behavioural principles. Let us examine
this behavioural theory in greater detail.

Where do Relational Frames Come From?

We have long known that organisms can respond to the formal relations between stimuli.
For example, many species can respond to the "dimmest" of several illuminated stimuli
(Reese, 1968). Such non-arbitrary relations are based on the formal properties of the relata
(i.e., one of the stimuli really is the dimmest). However, humans can respond in
accordance with relations that are controlled, not by the formal properties of the relata, but
by specific contextual cues.

Contextual control for relational responding becomes established during early language
training interactions. During these interactions, children are often presented with objects
and asked to repeat their names. This can be described as; see object A, then hear name B,
and say name B. Children are also taught to identify objects when they hear the
appropriate name. This may be described as; hear name B, then see object A. Initially,
each object-word and word-object relation is explicitly trained. However, when a child
 Relational Frame Theory 3

has been exposed to enough of this relational training, derived relational responding may
emerge. Suppose, for example, that a child with this history of naming is taught; "This is
your shirt". Contextual cues (such as the word "is", and the context of the social
interaction more generally) predict that if this object is a "shirt" (object A - name B), a
"shirt" is this object (name B - object A). Consequently, the child may now identify the
shirt when asked "Where is your shirt?" in the absence of differential reinforcement for
doing so. This derived, arbitrarily applicable relation is referred to as a "relational frame".
Thus, deriving relations is not genuinely novel, but is a type of generalized operant
behaviour. In other words, patterns of relational framing are brought under the control of
contextual cues (e.g., the word "is") through a process of differential reinforcement. That
is, to begin with, both elements of a relation are explicitly trained (e.g., "A is B" and "B is
A" are both reinforced). Only then can this history of reinforcement generalize so that a
derived relation emerges without reinforcement (e.g., if "X is Y" is reinforced, then "Y is
X" is derived). In effect, a well established principle of behaviour analysis, that of the
generalized operant, has been used by RFT to explain one of the key generative features
of human language.

Other types of stimulus relations that permeate human language may also be explained in
terms of generalized operant behaviour. Imagine, for instance, a young child who is taught
to respond to a range of questions such as "Which cup has more milk?" or "Which box
has more toys?" Given sufficient exposure to such questions and appropriate
reinforcement for answering them correctly, the relational response (e.g., between two
cups) will come under the control of cues other than the actual relative quantities (e.g., the
word "more"). When this occurs, the relational response can be arbitrarily applied to other
events, even when the formal properties of the related events do not occasion the
relational response. For example, a five pence piece is worth more than a two pence piece,
even though the former is smaller than the latter. This provides yet another example of the
way in which RFT explains advanced language and cognitive phenomena (e.g., a child's
understanding of financial value) in terms of a history of differential reinforcement that is
generalized to novel events.

Relational Frame Theory is a behavioural theory insofar as it draws together a number of

well established behavioural principles to explain many aspects of human language and
cognition. For example, RFT has drawn together the principles of respondent control (or
Pavlovian conditioning) and generalized operant behaviour to explain spontaneous and
apparently uncontrolled human anxiety (see also the subsequent paper). For illustrative
purposes, imagine a young child who hears that she is going on a "boat", and subsequently
experiences a terrible bout of sea sickness (i.e., the word "boat" becomes aversive via
Pavlovian conditioning). The child may then learn at school that a "Car Ferry" is a type of
boat. Later, on hearing that she is going on a car ferry, the child may show signs of
anxiety despite having had no direct experience with car ferries. This effect is based on
the respondently acquired function of "boat" and the derived relation between "boat" and
"car ferry". In effect, the child does not need to experience the possibly aversive
consequences of traveling on a car ferry in rough seas, in order to show signs of anxiety.
Several authors have combined behavioural principles in accordance with RFT (e.g.,
respondent conditioning and generalized operant behaviour) to account for a wide range
of complex psychological phenomena that have hitherto fallen outside the purview of
behaviour analysis, such as anxiety (Friman, Hayes & Wilson, 1997), depression (Hayes
& Wilson, 1993), rule following (Barnes, Healy & Hayes, in press), prejudice (Watt,
Keenan, Barnes & Cairns, 1991), self-awareness (Dymond & Barnes, 1995), self concept
 Relational Frame Theory 4

(Barnes, Lawlor, Smeets & Roche 1996), sexual arousal (Barnes & Roche, 1997b; Roche
& Barnes, 1997, 1998), spirituality and mysticism (Barnes & Roche 1997a; Hayes, 1984).