Language is full of words with more than one meaning (e.g., the word bat, which means nocturnal flying mammal as well as wooden club used in baseball), the so-called ambiguous words. It is estimated that approximately 44% of English words are ambiguous, a quantity that increases considerably for high-frequency words (Britton, Reference Britton1978). Given their abundance, understanding how ambiguous words are represented and processed by the human mind is necessary for a complete explanation of human language processing and representation.

A unique property of ambiguous words is that they show a particular one-to-many mapping between their spelling and their meanings. This property makes them very interesting and useful for psycholinguistic research, especially for studying how words are processed and represented. For example, research on semantic ambiguity has contributed to elucidating how semantics and orthography interact during word recognition (e.g., Balota, Ferraro, & Connor, Reference Balota, Ferraro, Connor and Schwanenflugel1991; Hino, Lupker, & Pexman, Reference Hino, Lupker and Pexman2002) or how a previous context affects meaning activation during reading (e.g., Swinney, Reference Swinney and Simpson1991; Van Petten & Kutas, Reference Van Petten and Kutas1987). In addition, ambiguous words have also been employed in other fields of research. Just to name a few, they have been used to investigate implicit memory and false memories (Eich, Reference Eich1984; Hutchison & Balota, Reference Hutchison and Balota2005), depression (Hertel & El-Messidi, Reference Hertel and El-Messidi2006), personality disorders (Mathews, Richards, & Eysenck, Reference Mathews, Richards and Eysenck1989), or autism (Hala, Pexman, & Glenwright, Reference Hala, Pexman and Glenwright2007).

Focusing on word recognition research, a common finding is that ambiguous words are recognized faster than unambiguous words in the lexical decision task (LDT; e.g., Hino & Lupker, Reference Hino and Lupker1996; Jastrzembski, Reference Jastrzembski1981; Kellas, Ferraro, & Simpson, Reference Kellas, Ferraro and Simpson1988; Millis & Button, Reference Millis and Button1989). This processing advantage for ambiguous words in the LDT (called ambiguity advantage) has been a challenge for most of the models of word recognition, especially for classical models and parallel distributed processing (PDP) models. Classical models were unable to provide an explanation for this effect, because they assumed that the information related to the meaning(s) of a word (i.e., semantic information) should not affect word recognition. Thus, to account for the ambiguity advantage, some changes in these models were proposed (e.g., logogen models, Jastrzembski, Reference Jastrzembski1981; and serial-search models, Rubenstein, Garfield, & Millikan, Reference Rubenstein, Garfield and Millikan1970). For instance, it was suggested that ambiguous words might be represented by more than one lexical entry, each one connected to a distinct meaning of the word. Consequently, it would be more probable to quickly select one of the lexical entries of an ambiguous word than to select the single entry of an unambiguous word, resulting in a processing advantage for ambiguous words.

In contrast, PDP models were not able to account for the ambiguity advantage. PDP models predicted a processing disadvantage for ambiguous words, given the assumption that ambiguous words have an inconsistent one-to-many mapping between their orthographic representation (e.g., bark) and their semantic representations (e.g., the sharp cry of a dog and the outside covering of a tree). This inconsistent mapping would produce a slower semantic coding for ambiguous words, probably due to a process of competition between their multiple semantic representations (e.g., Azuma & Van Orden, Reference Azuma and Van Orden1997). Thus, to provide an explanation for the ambiguity advantage consistent with PDP models, some authors have proposed that recognition times in the LDT do not reflect semantic coding, but orthographic coding (e.g., Hino & Lupker, Reference Hino and Lupker1996). Accordingly, because ambiguous words have multiple semantic representations, they would benefit from a large semantic feedback from the different meaning representations to their orthographic representation during word recognition, speeding up the orthographic coding and so then recognition and response times.

Although the above-mentioned accounts of the ambiguity advantage assume that this phenomenon is a result of words having multiple meanings, some studies conducted during the last decade strongly challenged this assumption (e.g., Armstrong & Plaut, Reference Armstrong, Plaut, Carlson, Hölscher and Shipley2011; Klepousniotou & Baum, Reference Klepousniotou and Baum2007; Rodd, Gaskell, & Marslen-Wilson, Reference Rodd, Gaskell and Marslen-Wilson2002). Such studies examined words with multiple related meanings (i.e., polysemes; e.g., newspaper), words with multiple unrelated meanings (i.e., homonyms; e.g., bat), and unambiguous words by using the LDT. They reported a disadvantage for homonyms in comparison to unambiguous words, and an advantage for polysemes with respect to unambiguous words and to homonyms. Thus, these studies showed that the degree of relatedness between the different meanings of a word, rather than the number of meanings, facilitates word recognition. To account for this evidence, Rodd, Gaskell, and Marslen-Wilson (Reference Rodd, Gaskell and Marslen-Wilson2004) suggested that the multiple semantic representations of a word with unrelated meanings would compete during word processing, slowing down their recognition in the LDT; in contrast, the related semantic representations of a word with related meanings would facilitate the recognition of the word. However, it should be noted that some findings are at odds with this relatedness of meanings advantage. Several studies have found an advantage of the same magnitude for polysemes and homonyms in comparison to unambiguous words (e.g., Hino, Kusunose, & Lupker, Reference Hino, Kusunose and Lupker2010; Hino, Pexman, & Lupker, Reference Hino, Pexman and Lupker2006).

In view of these conflicting experimental findings, more research is needed for a complete understanding of how ambiguous words are processed and represented, as well as to identify the variables that influence the processes and mechanisms involved in their recognition. However, before conducting a study on semantic ambiguity, researchers have to face several important issues. Three of them are, in our view, particularly critical: (a) selecting a proper set of ambiguous words; (b) categorizing distinct types of ambiguous words, and (c) selecting a matched set of unambiguous words for comparison purposes. In order to aid researchers to address these issues, the aim of the present study is to provide a database of ambiguous and unambiguous Spanish words. In what follows, we will briefly address the above-mentioned critical issues, focusing on studies that have used homographs (i.e., words with the same spelling but more than one meaning; e.g., fair). Then, in the Methods section, we will explain in detail the development of the database.

Selecting an appropriate set of ambiguous words is essential for experimental research on semantic ambiguity. In order to do so, researchers have to determine whether a word is ambiguous or not. This task can be achieved by employing different approaches, either objective or subjective. For example, an objective method to categorize words as ambiguous or unambiguous consists in counting the number of entries of the words in the dictionary (e.g., Jastrzembski, Reference Jastrzembski1981; Rodd et al., Reference Rodd, Gaskell and Marslen-Wilson2002). Using this approach (i.e., the dictionary approach), words with more than one entry are usually classified as ambiguous, whereas words with only one entry are classified as unambiguous. Regarding subjective approaches, a widely used method is to ask participants to generate definitions or lexical associates for a list of words. Then, the definitions or associates are grouped together according to the meaning to which they refer (Nelson, McEvoy, Walling, & Wheeler, Reference Nelson, McEvoy, Walling and Wheeler1980; Twilley, Dixon, Taylor, & Clark, Reference Twilley, Dixon, Taylor and Clark1994). The expected outcome is that definitions or associates for ambiguous words, but not for unambiguous words, will refer to more than one meaning. For instance, in Nelson et al. (Reference Nelson, McEvoy, Walling and Wheeler1980) some of the associates generated for the ambiguous word bass were fish, trout, drum, fiddle, or guitar, which are related to their two distinct meanings: sea perch fish and bass guitar. Finally, another subjective approach is to ask participants to think about all the meanings of a string of letters and then choose a value in a 3-point scale: (0) the word has no meaning, (1) the word has one meaning, or (2) the word has more than one meaning (e.g., Kellas, et al., Reference Kellas, Ferraro and Simpson1988; Pexman, Hino, & Lupker, Reference Pexman, Hino and Lupker2004). With this approach, a quantification of number of meanings (NOM) is obtained: words with values close to 2 are classified as ambiguous, and words with values close to 1 are classified as unambiguous.

Each of the above-described methods has some strengths and weaknesses. Concerning objective measures, dictionaries are an exhaustive and standardized resource of word meanings. Furthermore, most of them list unrelated meanings in different entries, and related meanings under the same entry, facilitating not only the selection of ambiguous and unambiguous words but also the selection of homonyms (i.e., multiple entries) and polysemes (i.e., only one entry). Given that, dictionaries seem to be a very useful tool to easily select a set of experimental stimuli. However, a strong argument against employing dictionaries in semantic ambiguity research was pointed out by Gernsbacher (Reference Gernsbacher1984). She claimed that dictionary definitions should not be taken as a psychologically valid measure of meaning representation. She argued that speakers do not store in their memory all the meanings listed in a dictionary, but only a small sample of them. As an example, she showed that college professors could provide only 2 of the 30 dictionary definitions of the word gauge, 3 of the 15 dictionary definitions of the word fudge, and 1 of the 15 dictionary definitions of the word cadet. Apart from this criticism, another reason against using the dictionary approach is the fact that dictionaries evolve quite slowly (Lin & Ahrens, Reference Lin, Ahrens, Minett and Wang2005). Thus, there are plenty of outdated definitions and they do not capture the emerging and gradual changes on the meaning of words.

With respect to subjective measures, asking participants to provide associates or definitions for a set of words would provide a better picture of which meanings are actually represented in the speakers’ minds. The reliability of these subjective measures has been evidenced in the studies that have relied on them to select their materials. For instance, it has been demonstrated that associates of ambiguous words produce significant priming effects in LDT (i.e., a facilitation in response time to target words preceded by related [e.g., hit-ball] vs. unrelated [e.g., doctor-ball] prime words; e.g., Klepousniotou, Pike, Steinhauer, & Gracco, Reference Klepousniotou, Pike, Steinhauer and Gracco2012). Nevertheless, a significant limitation of using associates is that to determine if a word is ambiguous or not as well as to count how many different meanings a particular word has, the responses of the participants have to be later examined and categorized by different judges. This task is time consuming, apart from being prone to errors and disagreements, because sometimes it is difficult to determine the meaning to which a particular response refers. For example, it is not clear if the word land, produced as an associate of the ambiguous word plane, is related to the airplane meaning or to the flat surface meaning. A similar case occurs when it has to be determined if two definitions provided for a word refer to exactly the same meaning, or instead, if they refer to distinct meanings. Consequently, this limitation may affect the ambiguous/unambiguous categorization of a word and the process of counting the number of different meanings. Finally, asking participants to assess the NOM a word has is a straightforward way to known if a word is ambiguous or not for speakers. In addition, it provides a numerical score of the word's ambiguity (i.e., a NOM rating), which facilitates the process of selecting stimuli for an experiment. However, there are two significant weaknesses of this approach. On the one hand, the criterion used by respondents to make their ratings is unknown. On the other hand, because ambiguous words can widely vary in NOM, this approach overlooks the differences between words with few meanings and words with many different meanings (Lin & Ahrens, Reference Lin, Ahrens, Minett and Wang2005). Given the above-mentioned advantages and limitations of each approach, we considered it convenient to include several different ambiguity measures in the present database, both objective and subjective. Therefore, the database provides the number of dictionary definitions, the number of dictionary senses, NOM ratings, and an ambiguous/unambiguous categorization made by judges on the basis of word associates.

The second critical issue to face when conducting research on semantic ambiguity is how to categorize different types of ambiguous words. These words vary along several dimensions. For example, they vary in NOM (few vs. many meanings), in meaning frequency/dominance (balanced vs. unbalanced meanings), and in the degree of relatedness between their meanings (related vs. unrelated meanings). Dimensions such as NOM or meaning frequency/dominance have been extensively studied in the past (for a review, see Simpson, Reference Simpson and Gernsbacher1994); in contrast, relatedness of meanings (ROM) has only recently begun to gain experimental interest (for a review, see Eddington & Tokowicz, Reference Eddington and Tokowicz2015). It is noteworthy that the results from this line of research have suggested that ROM is a relevant variable that determines the experimental effects obtained with ambiguous words.

In order to establish how ROM modulates word processing, a psychologically valid measure of this variable is needed. Researchers have employed different approaches to obtain it. For example, some authors ask participants to rate how related two definitions of an ambiguous word are (Azuma, Reference Azuma1996; Durkin & Manning, Reference Durkin and Manning1989). Similarly, other scientists ask participants to judge how related two sentences are, each sentence containing the same ambiguous word but varying the context in which it appears, one context being related to one meaning, and the other related to another meaning (e.g., Do you have a goal in life? Liverpool won by a goal to nil; Panman, Reference Panman1982). These approaches provide a clear context for each word meaning, which forces participants to focus exclusively on a particular meaning when making their decisions. This facilitates the task for the participant and, at the same time, allows researchers to obtain a clean score of the relatedness between the meanings of an ambiguous word. However, the main limitation of these approaches concerns the selection of the meanings to be presented to the participant. If only a subset of all the meanings of an ambiguous word is presented, the ROM rating may be biased (e.g., in case of selecting only two meanings of a word with three meanings, when two of them are strongly related but that the third one is unrelated).

An alternative approach to obtain ROM ratings is by asking participants to choose the appropriate ROM value for an ambiguous word using a numerical scale (Hino et al., Reference Hino, Pexman and Lupker2006). In this procedure, participants are presented with a list of ambiguous words, each one followed by a numerical scale comprising values from unrelated meanings to related meanings. Then, participants are required to think of all the meanings of each ambiguous word and to judge the relatedness of these meanings by selecting one of the values of the scale. The advantages and limitations of this approach are quite similar to those of NOM ratings. Its main advantages are that it is a straightforward approach to know if a word is polyseme or homonym, and that it provides a numerical value, which could be useful for selecting stimuli for psychological experiments. Regarding its limitations, two should be highlighted: the criterion used by participants to make a ROM decision is unknown, and it is questionable that participants consider all the meanings they know for a word before making the decision. Taking into account the strengths and weaknesses of the different ROM approaches, we decided to include two different ROM measures in the database, each one obtained through a distinct method.

Finally, as previously stated, the third critical issue when conducting research on semantic ambiguity is to choose a proper set of unambiguous words for comparison purposes. This is especially crucial for word recognition experiments, in which researchers typically compare ambiguous and unambiguous words with respect to the time needed by participants to recognize them. As pointed out above, a common finding in LDT is that ambiguous words are recognized faster than unambiguous words (e.g., Hino & Lupker, Reference Hino and Lupker1996; Jastrzembski, Reference Jastrzembski1981; Lin & Ahrens, Reference Lin and Ahrens2010; Millis & Button, Reference Millis and Button1989). Thus, in order to ensure that this advantage is produced by words’ ambiguity, ambiguous and unambiguous words have to be matched on all the relevant variables that are known to affect word processing (e.g., word frequency, word length, familiarity, concreteness). This experimental control becomes even more important taking into account that several interactions between ambiguity and other variables have been reported. Namely, the ambiguity advantage has been observed for low-frequency ambiguous words, but not for high-frequency ambiguous words (e.g., Pexman et al., Reference Pexman, Hino and Lupker2004); for abstract, but not for concrete words (e.g., Tokowicz & Kroll, Reference Tokowicz and Kroll2007); and for neutral, but not for emotional words (Syssau & Laxén, Reference Syssau and Laxén2012). For that reason, in the present database we included ratings for several variables, such as emotional valence, emotional arousal, concreteness, familiarity and age of acquisition.

To sum up, researchers should consider several issues before conducting semantic ambiguity experiments. To address these issues, it is very relevant to have large sets of normed stimuli. There are many databases for English ambiguous words (Azuma, Reference Azuma1996; Durkin & Manning, Reference Durkin and Manning1989; Ferraro & Kellas, Reference Ferraro and Kellas1990; Gawlick-Grendell & Woltz, Reference Gawlick-Grendell and Woltz1994; Gee & Harris, Reference Gee and Harris2010; Gorfein, Viviani, & Leddo, Reference Gorfein, Viviani and Leddo1982; Griffin, Reference Griffin1999; Nelson et al., Reference Nelson, McEvoy, Walling and Wheeler1980; Nickerson & Cartwright, Reference Nickerson and Cartwright1984; Panman, Reference Panman1982; Twilley et al., Reference Twilley, Dixon, Taylor and Clark1994; Wollen, Cox, Coahran, Shea, & Kirby, Reference Wollen, Cox, Coahran, Shea and Kirby1980). However, the number of available norms in other languages is much lower. Focusing on Spanish, only three normative studies have been published to date (Domínguez, Cuetos, & de Vega, Reference Domínguez, Cuetos and de Vega2001; Estevez, Reference Estevez1991; Gómez-Veiga, Carriedo López, Rucián Gallego, & Vila Cháves, Reference Gómez Veiga, Carriedo López, Rucián Gallego and Vila Cháves2010). These databases are described in detail in what follows.

The norms of Dominguez et al. (Reference Domínguez, Cuetos and de Vega2001) consist of 100 polysemous words. In this study, a sample of undergraduate students was asked to retrieve the different meanings of each word and to try to write three sentences with them. After that, the authors categorized the sentences according to the meaning to which the ambiguous word referred, and then they counted the number of distinct meanings. The database includes, for each word, the NOM, the number of responses assigned to each meaning, and the number of dictionary definitions. The database of Estevez (Reference Estevez1991) consists of 152 homonymous words and 61 polysemous words. A sample of 104 participants was asked to write definitions for the homonyms, and 96 participants were asked to write sentences referring to the different meanings of the polysemes. All of them were undergraduate students living in the Canary Islands. The responses were categorized according to the most common meanings listed for each word in the Spanish Language Dictionary published by the Real Academia Española (RAE, 1984). The data set provides, for each word, the NOM, the percentage of responses for each meaning, and the number of dictionary definitions. Finally, the norms of Gomez-Veiga et al. (Reference Gómez Veiga, Carriedo López, Rucián Gallego and Vila Cháves2010) contain 113 ambiguous words, which were evaluated by adults (ranging from 19 to 52 years old) and children (10 and 11 years old). Participants were asked to write down all the meanings they knew for each word in the same order that they retrieved them. The database includes the NOM of each word, the percentage of responses assigned to each meaning, and the familiarity of the word.

In light of the above, our opinion is that the available Spanish databases of ambiguous words lack some relevant information. Namely, (a) they only include one subjective measure of ambiguity, (b) they do not provide subjective ROM measures, and (c) none of them has a set of unambiguous words. Thus, the aim of the present study was to construct a normative database of ambiguous and unambiguous Spanish words that overcomes some of the limitations of the available Spanish databases. The present database is made up of 530 words. Four measures of ambiguity were obtained: ambiguity categorization based on lexical associates, subjective ratings on NOM, number of dictionary entries, and number of dictionary senses. In addition, we collected two subjective ROM measures that can aid researchers in selecting words to study the polysemy/homonymy distinction. Finally, subjective ratings for some relevant variables known to affect word processing were obtained in order to help researchers with the control of potentially confounding variables in ambiguity studies.

RESULTS

The complete database can be downloaded from http://psico.fcep.urv.cat/exp/files/haro_et_al_database_2016.xls. A summary of the stimulus characteristics of the database is shown in Table 1. In what follows, we will provide some reliability and validity data for the measures included in the database, as well as the main results of the analyses conducted to fully characterize the normed words.

Table 1. Descriptive stimulus characteristics for the 386 ambiguous words and the 144 unambiguous words (classified according to the judges’ categorization measure)

Note: NOM, Number of meanings; Ent, number of dictionary entries; Sen, number of dictionary senses; ROM, relatedness of meanings; AoA, age of acquisition; Fam, familiarity; Con, concreteness; Val, valence; Aro, arousal; Freq, frequency (number of occurrences per million); Lett, word length (in letters).

Data trimming

All the questionnaire versions for each measured variable were administered to at least 25 participants. Some respondents were rejected after applying a trimming procedure. To do so, we examined the data from the questionnaires in order to identify random or aberrant response patterns. For example, using graphical procedures for assessing person-fit, we discarded those participants with a pattern of responses with almost no variation, because this suggests that they used the same value of the scale for nearly all the words. Furthermore, we also computed a personal correlation coefficient between each participant's data and the mean, and we excluded those participants with negative values or values near to zero. Values near to zero would suggest that the participant responded randomly to the questionnaire, whereas negative values would suggest that the participant understood the scale in the opposite direction. As a result of applying this trimming procedure, the number of valid responses for each variable was, at least, 21 for NOM, 19 for ROM₁, 19 for ROM₂, 23 for emotional valence, 22 for emotional arousal, 24 for familiarity, and 22 for concreteness.

Reliability

We assessed the reliability of each variable through a split half intergroup procedure. This method consists of dividing the data from each questionnaire into two sets of scores, one set from even items and the other set from odd items. Then, the correlation between the two sets is calculated, giving an index of agreement between participants’ scores. Correlations were corrected using the Spearman–Brown formula. Mean intergroup correlation values were r = .85 for NOM ratings, r = .91 for ROM₁, r = .81 for ROM₂, r = .84 for emotional valence, r = .96 for emotional arousal, r = .98 for familiarity, r = .85 for concreteness, and r = .88 for age of acquisition. All the correlations were positive and significant (all ps < .001), supporting the reliability of the data.

Relationship between ambiguity measures

We obtained two objective measures of ambiguity (number of dictionary entries and number of dictionary senses) and two subjective measures of ambiguity (NOM ratings and the judges’ ambiguous/unambiguous categorization based on word associates). As can be seen in Table 2, objective and subjective measures were positively and significantly correlated. This is of relevance because it supports the validity of the subjective measures that we collected.

Table 2. Correlations between the measures of ambiguity and the ROM measures

Note: ROM, Relatedness of meanings; NOM, number of meanings.

**p < .01. ***p < .001.

In addition, it is important to note that ambiguous and unambiguous words categorized by judges differed in dictionary entries, t (528) = 4.82, p < .001, dictionary senses, t (528) = 10.76, p < .001, and NOM ratings, t (528) = 25.79, p < .001.

Relationship between ROM measures

Two ROM measures were collected: ROM₁ and ROM₂. ROM₁ ratings were obtained using a novel approach, whereas to get ROM₂ ratings, we followed the procedure used in a previous study (Hino et al., Reference Hino, Pexman and Lupker2006). The validity of our novel approach was supported by the high correlation between ROM₁ and ROM₂, r (528) = .60, p < .001.

Moreover, it should be noted that ROM₁ ratings were requested not only for ambiguous words but also for unambiguous words. Unambiguous words were included in ROM₁ questionnaires with the aim of obtaining additional support for the judges’ ambiguous/unambiguous categorization. Thus, we expected lower ROM₁ values (i.e., closer to the “unrelated meanings” value) for words classified as ambiguous than for words classified as unambiguous. Accordingly, ambiguous words had significantly lower ROM₁ ratings (M = 3.10) than unambiguous words (M = 7.19), t (528) = 29.69, p < .001. Likewise, this explains the negative correlation between ROM₁ and the ambiguity measures (see Table 2). In addition, ROM₂ was negatively correlated with both NOM ratings and number of dictionary meanings. Because ROM₂ ratings were only obtained for ambiguous words, this negative correlation indicates that ambiguous words with related meanings were considered by participants as having a lower NOM than ambiguous words with unrelated meanings.

Comparisons with other Spanish databases of ambiguous words

To assess the validity of the present database, we compared our ratings with those from other Spanish databases of ambiguity (Domínguez et al., Reference Domínguez, Cuetos and de Vega2001; Estevez, Reference Estevez1991; Gómez-Veiga et al., Reference Gómez Veiga, Carriedo López, Rucián Gallego and Vila Cháves2010), focusing on overlapping words across databases (range = 34–78 words). Because such norms only provide one subjective measure of ambiguity (i.e., NOM of the word) and no ROM measure, we could only examine the correlations between our NOM ratings and their measure of NOM. Significant correlations were observed between our data and those from Domínguez et al. (Reference Domínguez, Cuetos and de Vega2001), r (32) = .46, p = .007, as well as between our data and those from Gómez-Veiga et al. (Reference Gómez Veiga, Carriedo López, Rucián Gallego and Vila Cháves2010), r (57) = .39, p = .003. However, our ratings did not correlate with those of Estevez (Reference Estevez1991), r (76) = .09, p = .44. This may be due to methodological differences between studies. In the norms of Estévez (Reference Estevez1991), unlike in the other two studies, the responses of the participants were categorized according to the most common meanings provided by the dictionary. This is of relevance, because it might be possible that some responses related to other meanings were ignored, resulting in a bias in counting the meanings. In addition, Estevez relied on the 1984 edition of the RAE dictionary to categorize participants’ responses. There is a difference of 30 years between that version and the current RAE edition. Thus, it is likely that the most common meanings provided by the 1984 edition are slightly different from those included in the current edition. Finally, because the meanings of the words may evolve by the speakers’ use of language, it is also likely that the responses of participants in 1991 were somewhat different from those provided by respondents in the present study.

Conclusion

The aim of this study was to provide normative data for a set of ambiguous and unambiguous Spanish words. The resulting database is made up of 530 Spanish words rated on several subjective measures of ambiguity. It also contains dictionary measures of ambiguity (number of dictionary entries and number of dictionary senses), ROM measures, and values of some relevant lexicosemantic variables (i.e., concreteness, familiarity, emotional valence, arousal, and subjective age of acquisition).

The database includes two subjective measures of ambiguity and two subjective ROM measures. Although the two measures of ambiguity were obtained using distinct approaches (judges’ categorization based on word associates and NOM ratings), they were highly intercorrelated. In addition, both measures were correlated with number of senses as well as with number of entries in the dictionary. Furthermore, some significant correlations were observed between our NOM ratings and the ratings from other Spanish databases of ambiguous words. Altogether, these correlations provide support for the validity of the subjective measures of ambiguity. Similarly, ROM measures were obtained using distinct methods. A novel approach was used for collecting ROM₁ ratings, whereas ROM₂ ratings were obtained following the method employed in a previous study (Hino et al., Reference Hino, Pexman and Lupker2006). Despite the differences between both methods, the two ROM measures were highly intercorrelated, supporting the validity of the novel method. In light of this evidence, we consider that both ambiguity and ROM measures could be reliably used in semantic ambiguity research. In particular, they may be of value to assess the influence of the NOM and of the ROM on word recognition, helping to refine models of word recognition and to test their hypothesis about the processing and representation of ambiguous words.

Furthermore, the database includes some subjective ratings for concreteness, familiarity, emotional valence, arousal, and subjective age of acquisition that are not provided by any other available Spanish database. Namely, we collected new concreteness and familiarity ratings for 118 words, new emotional valence and arousal ratings for 339 words, and new age of acquisition ratings for 112 words. Given that, the present norms will be useful not only for researchers interested in studying semantic ambiguity but also for researchers looking for controlling such variables in their experiments.

In contrast, it should be highlighted that the present database might have some applications beyond psycholinguistic research. Ambiguous words have been used, among other situations, to study clinical populations. For example, by presenting homographs with both personal and impersonal meanings (e.g., close and console), Hertel and El-Messidi (Reference Hertel and El-Messidi2006) examined the tendency of people in depressed or dysphoric states to interpret ambiguous events as personally relevant. Similarly, semantic ambiguity has been employed as a tool to investigate the negative interpretation bias related to anxiety disorders. This has been done by examining the response of people with anxiety disorders when they were presented with ambiguous words having both threat and nonthreat meanings (e.g., patient; Hayes, Hirsch, Krebs, & Mathews, Reference Hayes, Hirsch, Krebs and Mathews2010). A further application can be found in the study of deficits in contextual processing in children with autism. For instance, Hala et al. (Reference Hala, Pexman and Glenwright2007) investigated whether children with autism are able to make use of the meaning of semantic primes to interpret the meaning of ambiguous words (e.g., pencil-lead). Thus, researchers interested in those or similar applications of semantic ambiguity could benefit from the present database.

In conclusion, this normative study overcomes some of the limitations of the published databases of Spanish ambiguous words: the scarcity of measures of ambiguity, the lack of subjective ROM measures, and the absence of a set of unambiguous words. As such, the database will be useful for researchers interested in studying semantic ambiguity as well as in their applications in different situations and/or populations. It will be especially valuable to assist researchers in categorizing ambiguous and unambiguous words, in categorizing ambiguous words that differ in the relatedness of their meanings, and in preventing any experimental confound due to uncontrolled variables.