The following is a description of the origin of responding under joint control in children. It is pictured in the diagrams of Horne & Lowe's 1996 JEAB article on the development of the naming relation. The point of this exercise is to illustrate that the conditions Horne & Lowe specify for the development of the naming relation are in fact exactly the conditions that produce joint control, and while Horne and Lowe may have identified the antecedent conditions correctly, it is argued here that these conditions have consequences other than those specified by Horne & Lowe. I have published another version of this argument in JEAB (1997) , 68, 244-247. That article was preceded by an article in JEAB (1996) 65, 252-255 (making points preliminary to these) as one of the commentaries accompanying the publication of Horne & Lowe.
Summary: Children's selection behavior comes under joint control, and thus capable of supporting novel bidirectional responding (i.e. what they call symmetry), only after three primary operant repertoires have been acquired and have interacted (at I). The first repertoire infants acquire, selection, consists of the non-mediated selection of stimuli in response to their names. Later (up to several months), children acquire the minimal vocal repertoire. This in turn allows the children to emit echoic behavior and to tact stimuli (the third relevant repertoire).
Later still (at II) these three repertoires (selection, echoic and tact) interact so that control over the selection of stimuli in response to their names changes from unmediated selection to selection under joint control.
Finally (at III), the generalizing characteristics of joint control come into play, and this source of control spreads to new words and objects; thereby producing bidirectional responding (word-object selection and object-word naming) with novel words and the objects they name.
2) Acquisition of the minimal vocal repertoire. Once the child can emit the sounds of English, s/he can repeat words heard (the echoic repertoire), and can name things (the tact repertoire). Here, on the left, in the echoic repertoire, the child says "dot" first as an echoic to /dot/, the sound of the word when spoken by another ("dot") . Also shown is the childs production of the word "dot" as a self-echoic response to the sound of their own emission of the word /dot/. To the right is a tact in which the child says "dot" in response to a dot.

(In Horne & Lowe's (1996) naming relation, illustrated
on the left, selection responses remain under the control of specific auditory
stimuli (e.g. /dot/). The accumulation of many such relations, they
assume, induces a higher-order generalizable relation. In contrast,
the development of generalization under joint control assumes that control
of the selection response shifts to an event common to all trials, and
thus the basis for generalized responding: the onset of joint control)

5) Under joint control, selection training with novel stimuli produces novel tacts. In a conditional discrimination task with novel stimuli, a child who has previously learned to select under joint control will rehearse sample names while learning which comparison to select (A). Each correct comparison selection produces reinforcement for emitting the rehearsed name (as an echoic) in the presence of that comparison (B). The topographies of the echoics thereby acquire tact functions (C). Thus, selection training produces tacts not directly trained: the bidirectional relation!
