The influence of training procedure and stimulus valence on the long-term maintenance of equivalence relations

Highlights

• The participants demonstrated the emergent stimulus-control following either DMTS and SMTS training.

• The performances of SMTS group on equivalence testing have declined 30 days later.

• The follow-up assessments showed the stability of equivalence-like performances among participants in the DMTS group.

• Overall the probability of class-consistent choices and derived transfer of function was enhanced for the classes composed of pictures of happy faces.

• Positive valence and DMTS procedure combined foster the long-term maintenance of equivalence classes.

Abstract

In the current research, we aimed at extending Silveira et al. (2016) results by verifying whether the long-term maintenance of the equivalence classes is influenced by stimulus valence and MTS training procedures. The delayed and simultaneous MTS were used to train two groups of participants in series of conditional relation trials involving pictures of humans’ faces expressing familiar emotions (A) and abstract forms (B, C, and D). The participants that demonstrated the emergence of novel BD and DB relations and class-consistent derived transfer of functions returned to the laboratory thirty days later. Follow-up assessments were given in which the probability of class-consistent responses was higher for the happy class only for participants exposed to DMTS training. This result shows that the maintenance of equivalence classes cannot be accounted for only in terms of the affective valence of the familiar stimulus. The affective valence of the happy faces may have yoked with the properties of DMTS, favoring the maintenance of the happy class. Thereby, we discussed the role of mediating verbal behavior evoked selectively by the pictures of happy faces appearing as samples that may have persisted during the delay interval as a possible mechanism underlying performances of participants trained in DMTS procedure.

Introduction

When stimulus-stimulus relations possess the properties of reflexivity, symmetry, and transitivity, the stimuli thus related comprise classes of equivalent stimuli that are substitutable for each other (see Sidman, 1994, for a comprehensive account). If a member of such an equivalence class acquires a specific function by direct training, this function usually transfers to the other class-members (e.g., de Rose et al., 1988; Dougher et al., 1994; Gandarela et al., 2020; Perez et al., 2017). Pre-experimental functions of a stimulus that becomes a member of an equivalence class also transfer to the other members (Barnes-Holmes et al., 2000).

Considering that facial expressions of basic emotions are stimuli with presumable innate functions (Ekman and Friesen, 1971; Izard, 1994), Bortoloti and colleagues (e.g., Bortoloti and de Rose, 2009; Bortoloti et al., 2013) demonstrated the transfer of functions of emotional valence to previously abstract stimuli. Bortoloti and colleagues used matching to sample procedure (MTS) to establish equivalence classes comprising sets of abstract shapes related to happy and angry facial expressions. A Semantic Differential (SD; de Almeida et al., 2014; Osgood et al., 1957) assessed the emotional valence of stimuli by ratings in scales anchored by opposite adjectives (e.g., sad-happy, tense-relaxed, rough-smooth, etc.). The SD ratings demonstrated that abstract stimuli acquired what may be called derived valence: the abstract stimulus equivalent to the happy facial expression acquired positive valence, whereas abstract stimuli equivalent to the angry facial expression acquired negative valence. These results were confirmed in studies that used other methods to assess the transfer of functions, such as the IRAP¹ (Bortoloti et al., 2019; Bortoloti and de Rose, 2012) and continuous measurements of brain potentials with an electroencephalogram (EEG; Bortoloti et al., 2014).

Despite the mutual substitutability of members of equivalence classes, research has shown that the relational strength of stimuli in equivalence classes (i.e., their relatedness) may vary as a function of parameters of training and testing, such as class- size, training structure, the delay between the presentation of sample and comparisons, motivational operations, etc. (e.g. Albright et al., 2019; Bortoloti and de Rose, 2009, 2011; Fields et al., 1995, 2020; Gomes et al., 2019). A recent study of Bortoloti et al. (2019; see also Bortoloti et al., 2020) showed that in equivalence classes comprising facial expressions of emotions, relatedness between stimuli may be enhanced by a happiness superiority effect (Kirita and Endo, 1995; Silveira et al., 2016). This effect was documented by increased transfer of valence functions from faces expressing happiness to equivalent abstract stimuli, when compared with transfer of valence from faces expressing neutral or negative emotions, such as fear, sadness, disgust, and anger (Bortoloti et al., 2019, 2020).

A study by Silveira et al. (2016) showed that the happiness superiority effect may also play a role in the maintenance of equivalence classes and the transfer of functions. This study used delayed matching to sample (DMTS) to teach conditional discriminations AB, AC, and CD, where the A stimuli were pictures of faces portraying neutrality, happiness, and anger, and the B, C, and D stimuli were abstract shapes. A total of 16 participants attained high levels of accuracy in tests for the derived BD and DB relations, which logically attest to the emergence of combined symmetry and transitivity of the directly taught relations (Sidman and Tailby, 1982). Silveira et al. demonstrated the establishment of three four-members ABCD equivalence classes (which will be referred to as, respectively, neutral, happy, and angry classes). SD ratings also demonstrated the transfer of the emotional valences from the meaningful A stimuli to the abstract D shapes. A follow-up assessment 30 days later showed differential maintenance of the classes. A total of 11 participants achieved high percentages of class-consistent choices in the B1D1 and D1B1 test trials (happy class), whereas eight showed high scores in the B2D2 and D2B2 test trials (neutral class), and only six showed comparable performance in the B3D3 and D3B3 test trials (angry class). A happiness superiority effect was also visible in the maintenance of derived valences, assessed via SD ratings obtained in the follow-up assessment. Evaluations of D1 showed that this stimulus was still similar in valence to the equivalent happy face, whereas ratings of D3 showed a marginal resemblance in valence to the equivalent angry face. The happiness superiority effect in the maintenance of equivalence and transfer of functions is consistent with studies showing that positive valence of facial expressions is strongly correlated with superior performance in recognition and long-term memory tasks (e.g., Liu et al., 2014; Nikitin and Freund, 2019). In summary, the happiness superiority effect has been evidenced in the study of Silveira et al. (2016) both in the enhanced relatedness among class members and the superior maintenance of the derived valences of the abstract stimuli equivalent to the happy faces.

The study of Silveira et al. (2016) taught stimulus relations via a DMTS procedure, in which a click on the sample removed it, and the comparison stimuli appeared 2 s later. DMTS has been found to enhance equivalence class formation as well as relatedness between stimuli (measured by the magnitude of transfer of functions). Studies by Arntzen (e.g., Arntzen, 2012; Arntzen et al., 2018) showed a higher probability of class formation after training with DMTS than with simultaneous matching (SMTS, in which the sample remained on the screen after the presentation of the comparisons). Bortoloti and de Rose (2009, 2012) showed that the magnitude of transfer of functions was higher after DMTS training. Also, de Almeida and de Rose (2015) showed that after relations between faces expressing emotions and arbitrary stimuli were reversed, the change in functions of the arbitrary stimuli were more pronounced after DMTS. Particularly, the study of Bortoloti and de Rose (2012) showed a happiness superiority effect: with DMTS training, transfer of functions was found for the abstract stimuli equivalent both to the happy and angry facial expressions, whereas with SMTS transfer of functions was found only for the abstract stimulus equivalent to the happy expression.

The aim of the current research was to examine the probability of long-term maintenance of the class-consistent responses as a function of training procedure (DMTS vs. SMTS) and stimulus valence (happy vs. angry; happy vs. neutral; angry vs. neutral; happy vs. neutral). The interaction between DMTS (as opposed to SMTS) and emotional expression found on the transfer results of Bortoloti and de Rose (2012) indicates that it is important to verify the interaction between delay and emotional expression also in the maintenance of equivalence and derived transfer. Therefore, we sought to extend the study of Silveira et al. (2016) by using two groups of participants, given SMTS or DMTS training, to establish three four-members equivalence classes comprised of the same pictures of emotional faces and arbitrary stimuli. Equivalence-like performances was measured immediately after training and approximately 30 days later.

In this regard, we used the mixed-effects logistic regression model (cf. DeHart and Kaplan, 2019; Young, 2018) for addressing the impact of these independent variables upon the probability of correct responses on BD and DB tests. Besides, a multivariate multiple regression model assessed the predicted probability of SD ratings made after the first exposure to MTS procedures (Day 1) and following MTS follow-up tests. Complementarily, the multiple regression model, compared the ratings for D stimuli made by the experimental groups (DMTS and SMTS) with the ratings of A stimuli made by a control group to provide quantitative estimations of the derived valences for D, in accordance with their class memberships.