Emotional Overtones of Behavior Analysis Terms in English and Five Other Languages

Abstract

It has been suggested that the language of behavior analysis is not always consumer-friendly, but the very limited empirical support for this claim comes from examining jargon in English. We consulted publicly available data sets to shed light on one specific aspect of the jargon problem: how non-English speakers may react emotionally to the technical vocabulary of behavior analysis. Previous research has suggested that English speakers may experience English technical terms as unpleasant. Here, we show that the same may apply when speakers of other languages (Egyptian Arabic, French, German, Brazilian Portuguese, and Spanish) encounter translated technical terms. Our results, although constrained by the availability of data for only a small sample of relevant terms, suggest that responses of English speakers to English terms may be a good predictor of emotional responding to translated terms. To our knowledge, this is the first empirical study to address international ramifications of a so-called marketing problem in behavior analysis. Our main purpose is to call attention to the need for cross-language and cross-cultural studies on factors that affect public perceptions and acceptance of behavior analysis.

A small but growing body of studies indicates that the “language of behavior analysis” might be an impediment to dissemination. Some of the words that comprise behavior analysis jargon may generate unpleasant emotions in nonexpert listeners, which could make and consumers leery of behavior analysts who speak in jargon (Critchfield, Becirevic, & Reed, 2017; Critchfield et al., 2017) or of interventions that described in technical language (Becirevic, Critchfield, & Reed, 2016; Jarmolowicz, Kahng, Invarsson, Goysovich, Heggemeyer, & Gregory, 2008; Rolider, Axelrod, & Van Houten, 1998; Witt, Moe, Gutkin, & Andrews, 1984). Much has been written about a “marketing problem” (Bailey, 1991) in which the language of behavior analysis may not be consumer-friendly (e.g., Doughty, Holloway, Shields, & Kennedy, 2012; Freedman, 2015; Foxx, 1996; Lindsley 1991), and empirical meat is now being put on the bones of this long-standing concern.

Virtually all empirical studies on this “marketing problem” have focused on English language behavior analysis jargon (Rolider & Axelrod 2005). Yet behavior analysis (ABA) has spread across the globe, and thus many interactions between professionals and consumers take place in languages other than English. To support these interactions, the Behavior Analyst Certification Board® provides glossaries of behavior analysis terms in 14 other languages (http://bacb.com/glossaries/). It is reasonable to wonder whether translated behavior analysis terms carry the same liabilities to dissemination as the originals appear to do in English.

Recently, in an effort to estimate how a number of behavior analysis technical terms might strike the typical listener, we have consulted published databases, or corpora, on how people respond emotionally to English words (Critchfield, Becirevic, et al., 2017; Critchfield, Doepke, et al., 2017). Psycholinguists created these corpora by asking presumably typical1 adults to rate their emotional reactions to thousands of words on a scale ranging from general pleasantness to general pleasantness. We searched the corpora for words that serve as behavior analysis terms, and treated mean ratings of these words as a predictor of normative consumer reactions to them. Consistent with concerns over language-based impediments to dissemination, many of the terms were perceived as unpleasant. We acknowledge that emotional responses to jargon are but one possible hurdle in the dissemination of behavior analysis, but observers have argued that it may be an important one (e.g., Bailey, 1991; Foxx, 1996; Lindsley, 1991).

Word-emotion corpora exist for several languages besides English, in theory making it possible to make similar estimates for those languages. The present investigation consults some of those corpora as a first step in assessing the consumer-friendliness of behavior analysis terms in languages other than English.

Study 1

The purpose of Study 1 was to compare word-emotion ratings of behavior analysis terms in English versus Spanish. We focused first on Spanish because behavior analysis has existed in Spanish-speaking nations for many decades (Ardila, 2006), and Spanish is the most-commonly spoken language other than English in the nation with the most behavior analysts (USA). Importantly, a large word-emotion corpus is available in Spanish.

Method

Data Sources

Data sources were two large, public-domain lists of words that have been rated for how they strike people emotionally. Stadthagen-Gonzalez, Imbault, Perez Sanchez, & Brysbaert, (2016) produced normative emotion ratings for about 14,000 Spanish words (hereafter referred to as the SG corpus). Rating procedures for the SG corpus were nearly identical to those of Warriner, Kuperman, & Brysbaert, (2013), who produced normative emotion ratings for about 14,000 English words (hereafter referred to as the Warriner corpus). This allows for direct comparisons between the two sets of ratings. Information on downloading the corpora can be found in the source articles.

Raters in the Warriner et al., (2013) and Stadthagen-Gonzalez et al., (2016) studies independently evaluated words, presented visually one at a time, on two dimensions of present interest. The first dimension, Valence, was rated on a scale of 1 = “Unhappy” (“Infeliz” in Spanish), implying negative emotion, to 9 = “Happy” (“Feliz” in Spanish), implying positive emotion2. For example, in creating their English corpus, Warriner et al., (2013) elaborated to raters that “Unhappy” was equivalent to “annoyed, unsatisfied, melancholic, despaired, or bored” and “Happy” was equivalent to “pleased, satisfied, contented, hopeful” (p. 1193). In keeping with the tenets of the source literature, Valence ratings will be described here as ranging from “Unpleasant” to “Pleasant.”

The second dimension, Arousal, was rated on a scale of 1 = “Calm” (“Tranquilo[a]” in Spanish) to 9 = Excited (“Exitado[a]” in Spanish). In instructions to participants, Warriner et al., (2013) elaborated that “Calm” was equivalent to “relaxed, sluggish, dull, sleepy, or unaroused” and “Excited” was equivalent to “stimulated, frenzied, jittery, wide awake, or aroused” (p. 1193). Arousal thus connotes intensity of behavioral activation and may be of interest to behavior analysts because of its conceptual similarity to Skinner’s (1953) account of emotions as motivating operations that make particular behaviors more or less reinforcing to engage in. For present purposes, therefore, as in our previous work, the Arousal scale will be described as ranging from “Not Motivating” to “Motivating” (Critchfield, Doepke, et al., 2017). Arousal is thought to interact multiplicatively with Valence to determine overall emotional responding to words (Warriner & Kuperman, 2015). For instance, between two highly pleasant words in the Warriner corpus, relaxation (Valence = 8.00) and orgasm (Valence = 8.10), the latter (Arousal = 7.19) would be predicted to be more impactful on a typical listener than the former (Arousal = 3.15).

Raters in Warriner et al., (2013) were 1827 workers in the Amazon Mechanical Turk® online data collection crowdsourcing platform (~ 60% female, with majority aged 30 or younger). Individual workers evaluated only a subset of the target words, and with a few exceptions (about 1% of words) resulting from data loss, ratings were obtained for each English word from at least 18 workers. Raters in Stadthagen-Gonzalez et al., (2016) were 512 college students (~ 80% female, M age = ~ 22 years) who participated through the Qualtrics® online survey platform. Individual raters evaluated only a subset of the target words, and ratings were obtained for each Spanish word from 20 individuals. In both studies, raters initially received practice applying the scales to words with known levels of valence and arousal based on previous research. Individuals whose ratings of these calibration words deviated substantially from previously established norms for were dropped from analysis.

It should be clear from the preceding that this method, in which words are presented in isolation, fails to capture the complexity of communication in everyday situations. However, emotional responses to words are thought to be to some degree context-independent, that is, separate from the linguistic structure and narrative flow of communication. For example, as Skinner (1957) observed, “Much of the emotional... behavior of the listener... has little to do with grammar or syntax. An obscene word has its effect regardless of its location or grammar” (p. 44). We propose that such emotional responses are an important adjunct to narrative relations and possibly an integral component of them (see Critchfield, Becirevic, et al., 2017). In the latter case, the relevant empirical literature suggests that a listener’s overall emotional response to narrative (i.e., a thematically related strings of words) can be predicted reasonably well by calculating the mean of normative emotional responses to the individual words of the narrative (e.g., Avey, Avolio, & Luthans, 2011; Floh, Koller, & Zauner, 2013; Petty, Schumann, Richman, & Strathman, 1993).

Selection of Terms and Assignment of Emotion Ratings

For 40 English behavior analysis technical terms for which word-emotion ratings are available in the Warriner et al., (2013) corpus (39 terms examined previously by Critchfield, Doepke, et al., 2017, plus chain, as in “behavior chain”), Spanish equivalents, when available, were identified from the BACB® translation glossary (October, 2014, edition, produced by a team headed by Javier Virues-Ortega3; download from http://bacb.com/wp-content/uploads/2015/07/Spanish-English-ABA-Glossary.pdf). For as many of the Spanish terms as possible, word-emotion ratings then were obtained from the SG corpus. This yielded a total of 35 English-Spanish pairs of terms.

Results and Discussion

Table 1 lists the terms and their normative emotion ratings (for English words, from the Warriner corpus; for Spanish words, from the SG corpus). Valence scores for the two languages were not significantly different for the two languages, with t(34) = 0.51, two-tailed p = 0.60. Figure 1 (left) shows, in scatter plot format, how English and Spanish Valence ratings corresponded for selected pairs of terms. English and Spanish Valence ratings covaried at r = + 0.93 (p < 0.0001). In terms of residuals, there was no consistent difference between English and Spanish ratings. Considering each term as a difference score (Spanish minus English), the mean difference was + 0.17, or about 2% of the rating scale range. Thus, in terms of perceived pleasantness, ratings of English terms were a good predictor of the ratings of Spanish equivalent terms.

Table 1.

Mean emotion ratings of behavior analysis terms in English and Spanish

English			Spanish
Term	Warriner Corpus		Term	SG Corpus		Dodds Corpus
	Valence	Arousal		Valence	Arousal	Valence
avoid	4.10	3.92	evitar	3.70	6.05	4.76
automatic	6.05	3.70	automático	5.60	4.45	6.48
behavior	5.28	4.70	conducta	5.10	5.55	5.32
chain	4.79	4.05	cadena	5.00	5.65	4.78
class	5.73	4.43	clase	5.60	5.40	5.98
conditioning	4.89	3.77	condicionamiento	5.85	4.80
consequence	3.86	4.31	consecuencia	4.25	6.70	4.12
contingency	4.67	3.22	contingencia	4.58	5.56
contingent	4.85	3.90	contingente	5.00	5.20
deprivation	2.58	4.57	privación	3.30	6.55
differential	4.00	3.63	diferencial	4.85	5.35
discrimination	2.45	5.62	discriminar	2.55	6.50
economy	3.64	5.32	economía	4.50	6.05	5.42
emit	4.80	3.96	emitir	5.60	5.65
environment	6.70	3.45	ambiente	5.95	4.40	6.70
equivalent	5.89	3.95	equivalente	6.00	4.40	6.08
escape	5.50	4.55	escape	5.50	6.05
extinction	3.10	5.00	extinción	2.80	7.25
function	5.55	4.10	function	5.50	4.55	6.32
functional	5.58	3.85	funcional	5.50	5.95	6.80
history	6.00	3.38	historia	6.28	4.65	6.44
negative	2.52	5.05	negativo	2.15	6.75	2.52
operation	2.94	5.71	operación	3.06	7.60	4.62
positive	7.57	5.50	positivo	8.00	4.45	7.68
punish	2.86	5.85	castigar	2.15	7.20
punishment	2.76	5.07	castigo	2.33	7.33	2.70
radical	4.36	5.27	radical	3.45	6.60	5.10
reinforce	5.53	3.96	reforzar	6.25	5.65
reinforcement	5.50	4.30	reforzamiento	6.10	4.85
reflex	5.24	4.70	reflejo	6.10	4.70	6.00
relation	5.84	4.36	relación	7.55	5.60	7.00
repertoire	5.47	3.27	repertorio	5.60	4.65
response	5.95	3.56	respuesta	6.03	5.53	6.80
token	5.63	3.09	fichas	5.65	4.45	5.42
verbal	5.29	3.76	verbal	6.00	5.00	5.96

For Valence the rating scale ranged from 1 (least pleasant) to 9 (most pleasant); for Arousal the scale ranged from 1 (least motivating) to 9 (most motivating). See text for details

Fig. 1.

Relationship of Valence (pleasantness) and Arousal (motivation) ratings for selected behavior analysis terms in English versus Spanish. See text for additional explanation

Figure 1 (right) shows that Spanish and English Arousal ratings also were positively correlated, but more modestly at r = + 0.67 (p < 0.0001). Spanish Arousal ratings were consistently higher than those for English, with a mean difference of + 1.32 (about 17% of rating scale range). This difference was statistically significant, with t(34) = 6.42, two-tailed p < 0.0001. Taken at face value, this effect implies that Spanish behavior analysis terms tend to have greater motivating properties than English terms—which, if correct, would be a matter of some concern. Recall that many behavior analysis terms are experienced as unpleasant (Critchfield, Doepke, et al., 2017), and that Valence and Arousal interact multiplicatively (Warriner & Kuperman, 2015). Consequently, a word with unpleasant emotional overtones may be experienced as especially unpleasant if it evokes strong Arousal, and our findings thus suggest the existence of a special linguistic marketing challenge for Spanish-speaking behavior analysts.

But other interpretations of Fig. 1 are possible. For example, a word’s emotional impact may be best understood, not in terms of its raw emotion rating, but rather in terms of how it compares to other words used by the same-language verbal community (Warriner & Kuperman, 2015). Normatively speaking, Arousal appears to be more muted in English (M = 4.2 in the Warriner corpus) than in Spanish (M = 5.3 in the SG corpus). The proper comparison, therefore, may be how an English term in relation to English words generally compares to its Spanish equivalent in relation to Spanish words generally. To illustrate, we divided the behavior analysis terms into four groups: Pleasant (positively-valenced) words of low and high arousal, and Unpleasant (negatively valenced) words of low and high arousal. Pleasant vs. Unpleasant and low vs. high arousal were defined in relation to means for the same-language corpus (rather than rating scale midpoints). White bars in Fig. 2 show the percentage of terms that fell into each category for English (left column) and Spanish (right column) variants. Black bars show the percentages of words overall in the two reference corpora that fell into these categories. In most cases, percentages were similar for behavior analysis terms and the overall same-language corpus, meaning that in most respects the terms we sampled were unremarkable by comparison to the overall language. There was one exception. A disproportionate percentage of negatively valenced Spanish behavior analysis terms showed high arousal. Thus, even after within-language normalizing, unpleasant behavior analysis terms may indeed be differentially impactful in Spanish. But our analysis is based on only a few dozen terms, and more data are needed, a point on which we will comment in the “General Discussion” section.

Fig. 2.

Normalized word emotion in English and Spanish. Words were categorized as pleasant vs. unpleasant via comparison to mean Valence rating of a same-language corpus. Words were categorized as high vs. low in Arousal via comparison of the mean Arousal rating of a same-language corpus. The figure compares patterns for behavior analysis terms to normative, same-language patterns. See text for more information

To summarize, Study 1 adds to a small but growing body of work suggesting that, consistent with worries about a “marketing problem,” some of the jargon of behavior analysis might be experienced by nonexperts as unpleasant (Critchfield, Becirevic, et al., 2017; Critchfield, Doepke, et al., 2017). The present study suggests that this phenomenon is not peculiar to English language terms, for we found that Spanish-native listeners react at least as negatively to many behavior analysis terms in Spanish as English native listeners do to behavior analysis terms in English. On the basis of this finding, we can project that behavior analysts who speak other languages would do well to heed the warnings of English-speaking behavior analysts about the importance of communicating pleasantly with nonexperts (e.g., Bailey, 1991; Foxx, 1996; Lindsley, 1991). Of course, the generality of our findings to languages other than Spanish remains to be tested, and this was the focus of Study 2.

Study 2

The goal of Study 2 was to explore the degree to which emotion ratings of English terms correspond to ratings of terms in four additional languages. We chose French, German, Portuguese, and Arabic because BACB® glossaries exist for these languages, suggesting that ABA services are being offered with some regularity in these languages. Note as well that for French, German, and Portuguese, the tradition of translating behavior analysis terms extends back well before the BACB® began offering glossaries (Azzi, Rocha, Silva, Bori, Fix, & Keller, 1963; Richelle 1960; Schaeffer 1960).

German is a member of the Germanic family of languages that includes English. French and Portuguese are members of the Romance family of languages that includes Spanish. The Germanic and Romance language families are branches of the larger meta-family of Indo-European languages that are thought to have developed from a common historical root language (Pereltsvaig, 2012). Moreover, cultural exchanges in Europe have resulted, over the centuries, in considerable borrowing among these languages. These are all reasons to expect that emotional reactions to many words may be similar across the three languages.

Arabic may be thought of as a possible control language. It is part of the Semitic language family that has unclear connections to historical Indo-European (Perelstvaig 2012). Arabic developed in areas of Northern Africa, many parts of which today differ culturally (e.g., in terms of religion and forms of government) from areas of Europe in which the other languages arose and are spoken. This provides reason to anticipate that emotional responses to Arabic and English terms might be less similar than those to Spanish and English terms. However, events such as European incursions into the Middle East during the Crusades and Medieval-era occupation of the Iberian Peninsula by Arabic-speaking African Muslims have contributed to some borrowing between Arabic and several European languages. This suggests the possibility that emotional responses to Arabic terms could parallel those of English equivalents.

Method

The main data source was a public-domain collection of word-emotion corpora that has been described by Dodds et al., (2015; hereafter called the Dodds corpora). We chose this data set because it contains native-speaker Valence ratings (though not Arousal ratings, which will not be discussed in Study 2) of around 10,000 words in each of several languages, all of which were obtained using identical procedures.

Participants who generated the four non-English corpora were 179 native speakers of French from France, 196 native speakers of German from Germany, 208 native speakers of Portuguese from Brazil4, and 185 native speakers of Arabic from Egypt5. The Dodds et al., (2015) report does not provide additional demographic information for these participants. The data collection procedures of Dodds et al., (2015) differed somewhat from those used to create the Warriner English and SG Spanish corpora. As in those instances, Valence was rated on a scale ranging from 1 = Unhappy to 9 = Happy, but scale anchors were face icons expressing the relevant emotions rather than numerals and verbal descriptors. Each word was rated independently by 50 individuals.

Our goal was to compare ratings of behavior analysis terms from the Dodds English corpus to those of equivalent terms in the other four Dodds corpora. However, we encountered a problem that we have seen often when consulting word-emotion corpora: Only 23 of the 35 terms of interest from Study 1 could be found in the Dodds English corpus. Furthermore, not all of these 23 words could be found in any single non-English corpus, so we faced the prospect of extremely small samples of term pairs for our cross-language comparisons. To maximize sample size, when a term of interest was absent from the Dodds English corpus, we used the relevant valence rating from the Warriner corpus of Study 1. In defense of this decision, we note that Warriner and Kuperman (2014) have documented considerable similarities between the Dodds and Warriner corpora. For the 23 terms that did appear in the Dodds corpus, ratings correlated with those in the Warriner corpus at r = + 0.93. This suggests that for present purposes the Warriner ratings are a reasonable substitute.

For the same 35 English behavior analysis technical terms for which Valence ratings were available, translated equivalent terms were sought in the BACB® French, German, and Portuguese glossaries. Valence ratings were obtained in this way for 14 French, 13 German, and 12 Portuguese terms.

Our evaluation of Arabic was constrained by our unfamiliarity with its alphabet, so to identify terms we did not consult the relevant BACB® glossary but instead relied entirely on a version of the Dodds corpus that provides English translations (which we did not attempt to corroborate). In this way, we obtained Valence ratings for 10 Arabic terms. For several of these terms, the corpus presents multiple ratings because several variations of a word may exist in Arabic for which there exist no separate English translations. In these cases, for purposes of analysis, we used the average of the multiple ratings.

Results and Discussion

Figure 3 summarizes the correspondence between Valence ratings of English and non-English terms. Table 2 shows raw ratings. Emotional responses of English native listeners to English terms closely paralleled those of non-English native listeners to terms in their own languages (for all correlations, p < 0.01). For French and Portuguese, Valence ratings correlated with English ratings about as strongly as two independent assessments of Valence in English (the Dodds and Warriner corpora, as noted above). Positive correlations also were evident for German and Arabic. This may be particularly noteworthy for Arabic because, as mentioned previously, there are reasons to expect differences between Arabic and the other languages examined here.

Fig. 3.

Relationship of Valence (pleasantness) ratings for selected behavior analysis terms in English versus four other languages. See text for additional explanation

Table 2.

Mean emotion ratings of behavior analysis terms in English and four other languages

English		French		German		Brazilian Portuguese		Egyptian Arabic
Term	Valencea	Term	Valence	Term	Valence	Term	Valence	Termb	Valence
avoid	3.14	eviter	4.34
automatic	6.00					automática			5.50
behavior	5.50	comportament	5.34	verhalten	5.34	comportamento	6.16		5.35
chain	5.16	chaîne	5.26	kette	5.66
class	5.52	classe	5.80	klasse	6.62				5.16
consequence	4.36	conséquence	4.66	konsequenz	4.60	consequencias	4.68
discrimination	(2.45)					discriminação	2.18
economy	4.38	economie	5.36			economia	5.76
environment	6.22	environnement	6.64	umfeld	5.70	ambiente	6.66		5.74
equivalent	5.50			entsprechend	5.32	equivalenta	5.70
escape	5.64			flucht	3.36				3.47
function	5.60	fonction	5.16	funktion	5.64	função	5.40		6.34
functional	6.38					funcional	6.34
history	5.84	historique	5.70			história	6.80		5.94
negative	2.42			negative	3.14				3.22
operation	3.72	opération	4.28	operation	3.70	operação	4.54		4.58
positive	7.80	positif	7.18	positive	7.54	positiva	7.68
punishment	2.00								2.94
radical	4.58	radical	4.06
relation	6.36	relation	6.52	beziehung	6.88
response	5.68			reaktion	5.88
token	(5.63)					ficha	5.66
verbal	5.56	verbale	5.38	verbaler

The Valence rating scale ranged from 1 (least pleasant) to 9 (most pleasant). English terms are listed only if a rating was available for at least one translated equivalent, and translated terms are listed only if a rating was available

^aIf no valence rating was available in the Dodds corpus, one was substituted from the Warriner corpus. These instances are shown in parentheses

^bTerm-equivalent ratings were determined based on word translations provided in the Dodds corpus. That is, raters reacted to the words printed in Arabic form, but unlike for the other languages we consulted a table of results in which Arabic words were already translated into English. Thus, terms in Arabic script are not reproduced here

General Discussion

Because behavior analysts must frequently interact with nonexperts, there is great interest in how people who might profit from the science and technology of behavior analysis react to the peculiar language of the discipline. Historically, this concern has been addressed mostly through anecdote and casual observation (e.g., Bailey 1991; Doughty et al., 2012; Freeman 2015; Foxx 1991, 1996; Lindsley 1991). The present investigation is part of an effort to rely instead on research to understand any “marketing problem” that may afflict behavior analysis.

The “marketing problem” is partly a problem in verbal behavior, one that, upon inspection, could prove to be quite complex even when isolated to a single language like English (see Becirevic, et al., 2016; Critchfield, Becirevic, et al., 2017; Jarmolowicz, et al., 2008; Witt, et al., 2008). The spread of behavior analysis around the world adds an additional layer of complexity that is ignored in the mostly English-focused discussions of the “marketing problem” that have been published to date. To our knowledge, the present investigation is the first to explore how behavior analysis terms, when translated into other languages, strike same-language listeners.

Our investigation was limited by the scope of available data sources, meaning that we were able to examine a relatively small number of behavior analysis terms in any one non-English language. Nevertheless, our results are striking in their consistency: For behavior analysis terms in five languages, emotion ratings corresponded strongly to those of the same terms in English. Based on this consistency of effect and the general finding that a majority of behavior analysis terms in English tend to elicit negative emotion (Critchfield, Doepke, et al., 2017), we propose that whatever word-emotion based “marketing problem” exists for behavior analysis in English-speaking verbal communities probably exists to some degree for many other verbal communities.

Currently, the scope of available word-emotion data makes gauging the magnitude of the “marketing problem” difficult, particularly for languages other than English. We have found that, for unknown reasons, words that serve as behavior analysis terms are better represented in English word-emotion corpora than in those for other languages. This highlights a useful contribution of the present investigation: When empirical guidance about the emotional impact of behavior analysis terms is missing for a language of interest, it may be helpful, as a point of departure at least, to extrapolate from English language data sets (e.g., see our previous English-specific findings: Critchfield, Becirevic, et al., 2017; Critchfield, Doepke, et al., 2017).

An obvious direction for future research is for behavior analysts to generate data sets that are specific to their own needs and interests. The procedures used to collect word-emotion ratings, which are somewhat laborious but not technically demanding (e.g., see Stadthagen-Gonzalez et al., 2016; Warriner et al., 2013), are possible for any interested behavior analyst to replicate. One essential goal is to obtain emotion ratings for a fuller range of behavior analysis terms than we have examined so far, something that could be accomplished for any language of interest.

It is important to note that the “marketing problem” does not simply span different languages. Within a single language, behavior analysts interact with many types of nonexperts, each of which potentially constitutes an independent verbal sub-community. Existing word-emotion research ignores this complexity, treating the “verbal community” as an omnibus entity. But there is no guarantee that different types of individuals who speak the same language will react the same way to behavior analysis terms. Research is needed to map the similarities and differences among sub-communities. We illustrate by noting that the same language is not spoken identically in all locations (compare Spanish in Spain vs. Mexico vs. the Philippines). Drawing upon 236 Mexican natives, Dodds et al., (2015) produced a corpus of emotion ratings for Spanish words that can be compared to those in the SG corpus that was compiled in Spain. We identified 23 pairs of same-word, different-nation Spanish ratings (Dodds Spanish ratings are shown in Table 1); Figure 4 summarizes their correspondence (r = + 0.91 overall). This finding shows that verbal sub-communities that share a language will not always differ systematically in emotional responding to behavior analysis terms; but this outcome can hardly be counted upon, and only new research can reveal divergences that do exist.

Fig. 4.

Relationship of Valence (pleasantness) ratings for selected behavior analysis terms in Spanish, as evaluated by listeners in Spain vs. Mexico. See text for more information

Figure 4 does contain possible clues of between-location divergences in emotional responding to terms. A dashed line in the figure shows strict equality between valence ratings in Spain vs. Mexico. For a term like radical, the valence rating fell rather far from this line, in this case suggesting that radical is, for some reason, experienced as more pleasant in Mexico than in Spain. Each such discrepancy is an opportunity for a verbal functional analysis to determine what aspects of culture, local history, or verbal traditions are responsible for it. The results of such an analysis might prove useful in guiding the development of consumer-friendly ways of communicating.

Nationality, of course, is only one possible line of fracture for verbal sub-communities. Listeners differ in terms of gender, age, type and amount of education, professional and personal experiences, and circumstances that bring them into contact with behavior analysts. Here, we take the liberty of relating a potentially informative anecdote. Both of the present authors have extensive experience teaching behavior analysis to preservice teachers-in-training and, although we have not collected supporting data, we perceive that special education students often are more receptive to, and more quickly able to master, behavioral jargon than students in other specializations (e.g., elementary education). If correct, this observation illustrates that, at a practical level, the “marketing problem” will be easiest to confront when something is known about how members of many different sub-communities respond to behavioral jargon, so that behavior analysts may better tailor their verbal behavior to the needs of different types of listeners. As we have pointed out elsewhere, “common sense” is a fallible guide to this process. Data serve better (Critchfield, Doepke, et al., 2017).

We do not, of course, mean to imply that word-emotion data can tell the entire story of how behavior analysis jargon affects nonexperts. Rather, for a topic that has generated far more talk than research, these data are a convenient place to start in evaluating possible effects of interest. We say “possible” effects because the ultimate interest is not on how people feel, or what they say about how they feel, but rather what they do in practical settings. For example, following an interaction with a behavior analyst, will they seek out services, and if asked to help implement, do so effectively? Will they recommend that particular behavior analyst and those particular services to others? The available research is nearly moot on these points6, defining yet another essential direction for future investigation. Until the needed data are available, word-emotion corpora provide some basis for anticipating how consumers might react to jargon in practice settings. In this regard, we suggest that word-emotion data be viewed as similar, in one key respect, to social validity assessments in which consumers are asked whether an intervention was found to be acceptable. Their answers are taken as a face-valid predictor of whether, in the future, the consumer is likely to seek out similar services and recommend them to acquaintances (e.g., Wolf, 1978), but data directly relevant to this assumption almost always are lacking (Carr et al., 1999). In other words, social validity assessments, though uncorroborated, are embraced because they are easy to obtain and alternative sources of relevant information often are difficult to come by (Wolf, 1978). Word-emotion data are simply one way to evaluate the “social validity” of behavior analysis jargon, and our purpose in the present report is to suggest that when social validity of interest it is wise to consider that not all consumers may be alike (e.g., Fong et al., 2016).

Compliance with Ethical Standards

Ethical Approval

This article does not contain any studies with human participants or animals performed by any of the authors and therefore is not subject to Institutional Review Board oversight.

Conflict of Interest

The authors declare that they have no conflict of interest.