Research in psycholinguistics requires that the features of experimental stimuli be identified correctly and manipulated systematically in order to avoid potentially biased results. Cognitive research designs, as in studies of semantic memory (Smith, Reference Smith and Estes1978; Tulving, Reference Tulving, Tulving and Donaldson1972), require that selection of verbal and visual stimuli be controlled in order to define the research question with precision and clarity. One method of doing so involves the use of normative studies, such as those examining frequency of word use (Alameda & Cuetos, Reference Alameda and Cuetos1995; Alonso, Fernandez, & Díez, Reference Alonso, Fernandez and Díez2011; Duchon, Perea, Sebastián-Gallés, Martí, & Carreiras, Reference Duchon, Perea, Sebastián-Gallés, Martí and Carreiras2013; Justicia, Reference Justicia1995; Manzano, Piñeiro, & Pereira, Reference Manzano, Piñeiro and Pereira1997), which plays a vital role in word recall. In studies of lexical relations, however, Word Association Norms are typically used (Alario & Ferrand, Reference Alario and Ferrand1998; Deyne & Storms, Reference Deyne and Storms2008; Fernandez, Díez, & Alonso, Reference Fernandez, Díez and Alonso2010; Macizo, Gómez-Ariza, & Bajo, Reference Macizo, Gómez-Ariza and Bajo2000; Schulte, Borgwaldt, & Jauch, Reference Schulte, Borgwaldt and Jauch2012).

Word Association Norms (WANs) are databases, usually comprising two essential elements: stimulus words (SWs) and response words (RWs). WANs are generally created through free association tasks (Coronges, Stacy, & Valente, Reference Coronges, Stacy and Valente2007; Deyne & Storms, Reference Deyne and Storms2008; Salles, Holderbaum, & Machado, Reference Salles, Holderbaum and Machado2009), in which participants are asked to indicate, either orally or in writing, the first word that comes to mind when they read or hear a stimulus word (e.g., dog–cat). WANs have several objectives: to determine what lexical relations are most frequently produced by speakers of a given language variety; to measure the strength of association between words; and to ascertain whether or not two words are directly related. They can also be used to explore the different relations that are produced at different ages (Hirsh & Tree, Reference Hirsh and Tree2001). WAN databases facilitate the process of selecting lexical stimuli in research by providing indices and measures of word sets, including the associative strength of the first associate, the number of associates, and the number of idiosyncratic responses (Callejas, Correa, & Lupiáñez, Reference Callejas, Correa and Lupiáñez2003; Macizo, Gómez-Ariza, & Bajo, Reference Macizo, Gómez-Ariza and Bajo2000; Nelson & Bajo, Reference Nelson and Bajo1985). WANs are often employed in the design of semantic memory experiments, such as those involving lexical decision or naming tasks, to ascertain which words have activating or inhibiting effects (Alario, Segui, & Ferrand, Reference Alario, Segui and Ferrand2000; Ferrand & New, Reference Ferrand and New2003; Hutchinson, Balota, Cortese, & Watson, Reference Hutchinson, Balota, Cortese and Watson2008; Perea & Rosa, Reference Perea and Rosa2002). WANs are also used in studies of false recall and false recognition, in which lists of associated words are compiled to explore false memory generation (Cadavid, Beato, & Fernández, Reference Cadavid, Beato and Fernández2012; Gallo & Roediger, Reference Gallo and Roediger2002; Roediger & McDermott, Reference Roediger and McDermott1995). The frequency of a response word given for particular stimulus word indicates the associative strength of the two words in semantic memory (Nelson, McEvoy, & Dennis, Reference Nelson, McEvoy and Dennis2000; Salles et al., Reference Salles, Holderbaum, Becker, Rodrigues, Liedtke, Zibetti and Piccoli2008). By analyzing word associations, researchers obtain representative and reliable numerical data on the relations between words for speakers of the same language variety, which can be used to measure lexical effects such as priming. Lexical priming is generally defined as an effect whereby a word (e.g., “bee”) is processed faster if the preceding word is semantically and/or associatively related (e.g., “honey”) than if it is unrelated (Alario et al., Reference Alario, Segui and Ferrand2000; Ferrand & New, Reference Ferrand and New2003; Hutchinson et al., Reference Hutchinson, Balota, Cortese and Watson2008; Meyer & Schvaneveldt, Reference Meyer and Schvaneveldt1971; Perea & Rosa, Reference Perea and Rosa2002). Accordingly, priming effects depend on stored information about the relation between two words.

Stimulus words for priming tasks are generally drawn from WAN corpora so that stimulus selection is systematic and based on associative frequencies in a sample that speaks the same linguistic variety. The criterion of associative strength will depend on the specific research goals (Alario et al., Reference Alario, Segui and Ferrand2000; Perea & Rosa, Reference Perea and Rosa2002). Selecting related or unrelated word pairs using WANs is one of the best ways to predict facilitation or inhibition effects in a population that speaks the same language variety. WANs also allow researchers to explore the type of lexical relation between stimulus words (SWs) and response words (RWs), and to measure specific effects: semantic, associative, or combined (Alario et al., Reference Alario, Segui and Ferrand2000). This sort of task has been applied to populations of adults (Lucas, Reference Lucas2000; McRae & Boisvert, Reference McRae and Boisvert1998; Meyer & Schvaneveldt, Reference Meyer and Schvaneveldt1971), school-age children (Nation & Snowling, Reference Nation and Snowling1999; Schvaneveldt, Ackerman, & Semlear, Reference Schvaneveldt, Ackerman and Semlear1977), and even clinical populations with amnesia (Schacter, Reference Schacter1985; Shimamura & Squir, Reference Shimamura and Squire1984) or neurological disorders such as multiple sclerosis, which can lead to impairments in language processing (Hagoort, Brown, & Swaab, Reference Hagoort, Brown and Swaab1996; Ukkonen, Vahvelainen, Hämäläinen, Dastidar, & Elovaara, Reference Ukkonen, Vahvelainen, Hämäläinen, Dastidar and Elovaara2009). Recent experimental adaptations have provided evidence that lexical relations are formed in the second year of life (Arias-Trejo & Plunkett, Reference Arias-Trejo and Plunkett2009; Rämä, Sirri, & Serres, Reference Rämä, Sirri and Sarres2013).

English-language WANs have long been used to examine lexical effects. Kent and Rosanoff (Reference Kent and Rosanoff1910) were the first to attempt an English WAN; they used a list of 100 stimulus words on a sample of 100 adults. Other authors, including Palermo and Jenkins (Reference Palermo and Jenkins1964), later created larger WAN databases. Of the English-language WANs most widely used, three in particular are worth noting: The Edinburgh Associative Thesaurus (Kiss, Armstrong, Milroy, & Piper, Reference Kiss, Armstrong, Milroy and Piper1973) includes 8,400 stimulus words used with 100 participants; the Birkbeck Word Association Norms (Moss & Older, Reference Moss and Older1996) has 2,464 stimulus words that were presented to British adults; and the University of South Florida Association Norms (Nelson, McEvoy, & Schreiber, Reference Nelson, McEvoy and Schreiber2004), the largest WAN database in the English language, which includes 5,019 stimulus words and responses collected over more than 20 years. WAN databases have also been compiled in other languages, including German (Walde, Borgwaldt, & Jauch, Reference Walde, Borgwaldt and Jauch2008), Dutch (Deyne & Storms, Reference Deyne and Storms2008), and French (Alario & Ferrand, Reference Alario and Ferrand1998).

In Spanish, there is a WAN database of 5,819 stimulus words from the European Spanish dialect (Fernández et al., Reference Fernandez, Díez and Alonso2010) that were gathered from several sources, including 247 frequent words corresponding to the standardized drawings of Snodgrass and Vanderwart (Reference Snodgrass and Vanderwart1980); 664 from false recall norms created by Fernández et al., (Reference Fernandez, Díez and Alonso2010); 1,372 from a normative study by Algarabel, Sanmartín, García, and Espert (Reference Algarabel, Sanmartín, García and Espert1986); and 3,536 from various normative studies in Spanish and participants’ responses to associative norms presented in various studies (Fernández, Diez, Alonso, & Beato, Reference Fernández, Diez, Alonso and Beato2004; Fernández et al., Reference Fernandez, Díez and Alonso2010). Fernández et al. (Reference Fernandez, Díez and Alonso2010) report norms based on analysis of stimulus and response words, direct and backward associative strength, number of associates, and blank responses. Their database is available online and allows users to search for relationships in both directions: stimulus word → response word, and vice versa. There is also a Spanish-language corpus (Macizo et al., Reference Macizo, Gómez-Ariza and Bajo2000) that presents associative data for 58 words collected from a sample of 100 school children 8 to 13 years old. Macizo and his collaborators (2000) compared their results to word-association data in adults (Algarabel, Ruíz, & Sanmartín, Reference Algarabel, Ruíz and Sanmartín1988), taking into account associative strength, number of associates, and idiosyncratic responses, among other measures. They concluded that developmental age influences the organization of associative knowledge.

WANs reflect the language and culture of participants in the sample. Thus, norms obtained in one language cannot be directly extrapolated to another. Even within the same language, the features of a given stimulus may vary from one culture to the next due to factors such as frequency of use, word or spelling variants, pragmatic questions, and other considerations. As numerous authors have concluded (Macizo et al, Reference Macizo, Gómez-Ariza and Bajo2000; Manzano et al., 1997; Sanfeliu & Fernández, Reference Sanfeliu and Fernández1996), there is therefore a need to study different varieties of the same language within specific cultures.

Some normative studies have been conducted in Mexican Spanish. Aveleyra, Gómez, Ostrosky-Solís, Rigalt, and Cruz (Reference Aveleyra, Gómez, Ostrosky-Solís, Rigalt and Cruz1996), for example, aimed to standardize a set of visual stimuli to be used in semantic memory research. To date, however, there are no WANs for Mexican Spanish, restricting word selection to norms created for European Spanish. Given the lack of Association Norms in Mexican Spanish, the main objective of this study is to create such a database.

Results

A total of 67,743 response words (RWs) were collected from 585 applications of the test (Version A = 304, B = 281). Seven tests, in which participants responded to fewer than 60% of SWs (70 of the 117 SWs per list), were eliminated, leaving a total of 67,622 RWs (99.82%) for quantitative analysis. The average blank response rate for individual SWs was 3.07 %. Responses consisting of a nonsensical series of letters (e.g.,“djgfsd”) were also eliminated. RWs in English were eliminated if they merely copied or translated the stimulus word (e.g., vestido–dress), but if they referred to another word, they were left as written (e.g., computadora [computer]–mouse). Participants responded with English words to 0.22% of the SWs presented.

The response words (RWs) were stored in a database. Spelling and accent marks were later standardized, using the Diccionario de la Real Academia Española (RAE) and the Diccionario del Español de México (DEM). Any extraneous characters distorting the meaning of a response were eliminated (e.g., vestido– largopp → largo [dress–longpp → long]). Also, to avoid dividing associative strengths among variations of the same word (e.g., juguete [toy]–niño [boy], niña [girl], niños [boys/children], niñas [girls] → niño [boy/child]), words were grouped according to the criteria suggested by Goikoetxea (Reference Goikoetxea2000): grammatical variations by gender were combined under the masculine form (e.g., pelota [ball]–niño [boy], niña [girl] → niño [boy/child]), and variations in number were combined under the singular form (e.g., pelota [ball]–niños [boys/children], niñas [girls] → niño [boy/child]), unless the proportion of feminine or plural responses was significantly higher than the proportion of masculine or singular responses. The feminine form prevailed in 3 cases: falda–niña [skirt–girl]; muñeca–niña [doll–girl]; aretes–niña [earrings–girl], and the plural form in the following cases: lentes–ojos [lenses–eyes]; plumones–colores [markers–colors]; patines–ruedas [skates–wheels]; crayolas–colores [crayons–colors]; dientes–blancos [teeth–white]; and botas–zapatos [boots–shoes]. Variations in tense were grouped with the infinitive (e.g., pelota [ball]–jugaba [he/she played], jugar [to play] → jugar [to play]); however, where a verb could not be distinguished from a noun or an adjective (e.g., juego [I play vs. game]). Finally, response words with the same root were combined (e.g., abeja [bee]–piquete [sting], picadura [sting] → picadura [sting]), opting for the variation most frequently used according to Diccionario del Español de México [Spanish Dictionary of Mexico] (DEM) and Real Academia Española (RAE) (2001) [Spanish Royal Academy].

For each SW, we calculated the response frequency and eight different measures often reported in the literature to provide information about various factors, including the degree of association between SW and RW. These measures include Associative Strength of the First Associate (FA); Associative Strength of the Second Associate (SA); Sum of Associative Strength of first two Associates (SM); Difference in Associative Strength between the first two Associates (DF); Number of Different Associates (NA); Blank Responses (BLR); Idiosyncratic Responses (IR); and Cue Validity of the First Associate (CV). These measures were suggested in studies by Callejas et al. (Reference Callejas, Correa and Lupiáñez2003); Macizo, Gómez-Ariza, and Bajo (Reference Macizo, Gómez-Ariza and Bajo2000); and Nelson and Bajo (Reference Nelson and Bajo1985), except for cue validity of the first associate (CV), which came from McRae, Cree, Seidenberg, and McNorgan (Reference McRae, Cree, Seidenberg and McNorgan2005). Table 1 presents conceptual and operational definitions of each measure. These measures make it easier to consult and intrepret data, given that the number of RWs per SW differed after the sorting criteria were applied. They will also make it possible to systematically select verbal stimuli in future studies. Appendix 1 displays results for all 234 SWs, in alphabetical order, on these eight measures.

Table 1.

Conceptual and Operational Description of Percentage Measures

Table 2.

Contrasting Examples of Cue Validity (CV) Measure

$$CV{\rm{(}}{C_j}|{F_i}{\rm{)}}\, = \,{{CV({C_i}|{F_j})} \over {\sum {CV({C_i}|{F_k})} }}$$

Note: The above formula was modified from the one originally proposed by McRae, Cree, Seidenberg & McNorgan (Reference McRae, Cree, Seidenberg and McNorgan2005).

The Associative Strength of First Associate measure was used to classify RWs, in keeping with the proposal of Salles et al. (Reference Salles, Holderbaum, Becker, Rodrigues, Liedtke, Zibetti and Piccoli2008). Those authors proposed three numerical ranges to classify associative strength: an associate is considered weak if it receives less than 10% of total responses; medium between 10 and 24.99%; and strong with 25% or more. Of the WANs presented here, we found 113 strong First Associates (range = 25–75.66 %), 119 medium First Associates (range = 10.10–24.9%), and 2 weak First Associates: botas–zapatos [boots–shoes]; and ventana–aire [window–air] (range = 7.63–8.67 %). Some examples of word pairs with strong associatiation include cuna–bebé [crib– baby], abeja–miel [bee–honey], suéter–frío [sweater–cold]; examples of medium association include sol–calor [sun–heat], puerta–abrir [door–open], and uvas–vino [grapes–wine].

The average Number of Associates (NA) is the total number of different responses generated by each of the 234 SWs. The mean NA per SW was 56.97 (range = 25–98). The SW that generated the fewest different responses was carriola [baby carriage] (n = 25), while computadora [computer] generated the most (n = 98).

Finally, a descriptive analysis of the Cue Validity of First Associate (CV) measure revealed that 112 (48 %) of 234 SWs had a CV of 1.0. That is, the RW produced as First Associate was a response exclusively to that word (e.g., the RW miel [honey] was the First Associate only when the SW was abeja [bee]). The remaining 122 SWs had CVs between 0.09 and 0.50, indicating that certain responses were produced as First Associates for multiple SWs (e.g., dormir [sleep] is the first associate for the SWs cama [bed], pijama [pajamas], and almohada [pillow]). Using these data, researchers will be able to select word pairs as a function of their associative strength or internal consistency.