Research in psychological science is showing an increasing sensitivity for aspects related to the adaptive function of cognitive mechanisms and operations, expanding both the range of factors that are of relevance in explanations of human behavior and the spectrum of questions that deserve careful investigation by cognitive neuroscientists. As a result, current topics of interest include examination of adaptive features in areas such as perception (Witt, Proffitt, & Epstein, 2004), emotion (Al-Shawaf, Conroy-Beam, Asao, & Buss, 2016), language (Hurford, 2007), or thinking (Haselton et al., 2009). Recent approaches to episodic memory have also adopted an evolutionary perspective (see an early proposal by Glenberg, 1997), with an interesting line focusing on aspects that emphasize the adaptive function of memory in terms of its survival utility (Nairne, Pandeirada, & Fernandes, 2017). The aim of the present study was to contribute to the development of methodological tools that can be of assistance in designing experimental paradigms aimed at further exploring and characterizing survival-related aspects of memory in the domains of episodic and semantic processing.

Seminal work conducted by Nairne and collaborators a decade ago (e.g., Nairne, Pandeirada, & Thompson, 2008; Nairne, Thompson, & Pandeirada, 2007) convincingly demonstrated that processing episodic information with a focus on survival features tends to result in superior memory for that information. Their experimental procedure of choice, since then successfully applied by many other researchers, is known as the survival-processing paradigm. In a typical experiment using this procedure, participants may be asked to rate a series of words in a number of dimensions, the critical one being a survival-relevant dimension (e.g., securing food). The result is that, in a subsequent memory test, words initially rated for survival value are better remembered than words rated with respect to other control dimensions, such as rating the stimuli for pleasantness or personal relevance. Similar results are found when the procedure involves considering the relevance of the words to survival versus non-survival imagined scenarios (e.g., finding shelter from predators versus moving personal belongings to a new home) or, as shown by Fernandes, Pandeirada, Soares, and Nairne (2017), when the words refer to objects that are encoded as potentially dangerous because of contextual interactions with contaminant agents (e.g., being touched by a sick person).

This type of survival-processing paradigm has proven to be powerful in terms of reliable findings, with results widely replicated with varied participants (Aslan & Bäuml, 2012), and diverse materials (Otgaar, Smeets, & van Bergen, 2010), and when comparisons involve a variety of control scenarios (Bell, Röer, & Buchner, 2013). The abundance of empirical findings has been accompanied by a search for explanations and has led to theoretical discussions that, while generally agreeing on the adaptive nature of the effect, tend to disagree on the specific processes involved in its production. Proposals have been advanced that emphasize the role of well-known mechanisms such as attentional bias, deep processing, elaboration, encoding variability, and self-reference (see Nairne & Pandeirada, 2016). However, as pointed in a recent meta-analytic review (Scofield, Buchanan, & Kostic, 2018), research on the topic has not yet provided conclusive evidence that would allow for a consensus in regard to the explanation of the phenomenon.

At this point, advancing the understanding of survival effects on memory might benefit not only from posing new questions and obtaining new data based on the survival-processing paradigm, but also from convergent evidence obtained with alternative empirical procedures and materials. One illustrative case in memory research is how the important issue of memory distortion has been empirically addressed, drawing evidence from a number of paradigms, from the ecologically oriented (e.g., Loftus & Palmer, 1974) to the more controlled laboratory setting (Roediger & McDermott, 1995). Perhaps a more relevant example here is the case of research on the effects of emotion on memory, where research in the human domain has made use of diverse empirical methodologies and stimuli. Using a procedure that is certainly similar to the survival-processing paradigm, researchers have been able to assign emotional value to relatively neutral stimuli by having them processed in the context of an emotionally charged scenario, leading to superior memory for emotionally processed items in standard memory tests (e.g., Cahill & McGaugh, 1995). But valuable information about the mechanisms behind the memory-modulating effects of emotion has also been obtained by making use of materials with intrinsic emotional value, demonstrating that stimuli that are in themselves high in emotional content or tone are better remembered in a variety of experimental situations and episodic memory tasks (e.g., Kensinger & Corkin, 2003).

In exploring the effects of survival on memory, little has attention has been focused on the intrinsic properties of stimuli. In a notable exception, some researchers have recently used a specific type of stimuli, animated concepts, to show that words that represent entities that have the characteristic of being animate, assumedly a survival-related feature, are better remembered, even in the absence of explicit instructions to the participants to engage in survival-centered encoding (Bonin, Gelin, & Bugaiska, 2014; Nairne, VanArsdall, Pandeirada, Cogdill, & LeBreton, 2013). While the animacy status of an entity can rationally be argued to reflect its survival value, an optimal approach for characterizing the survival value of specific stimuli would be to assess the relevance of the property more directly, and by means of empirically validated procedures; for example, characterizing concepts in terms of concrete survival dimensions. Having this type of descriptive information about words or images, for instance, would undoubtedly open the way for new manipulations and potentially contribute to further understanding survival effects on memory and cognition. For example, Howe and Derbish (2010) used associative word lists likely to be high in survival value to show how words scoring high were more susceptible to distortion in an experimental false-memory paradigm (the DRM procedure, Roediger & McDermott, 1995). While informative when used to test an adaptive property of survival-related information, the set of stimuli was small and limited in that it originated in a very reduced associative word pool, and in that the words were first selected as low or high in survival relevance on the basis of the researchers’ intuition (although, later, participants’ judgments largely agreed with the a priori established dichotomy). Similarly, a pilot study on a reduced set of 45 words was used by Butler, Kang, and Roediger (2009) to assess their survival value in a study showing that congruence between materials and processing task could strongly modulate the observed advantage of survival processing in episodic memory situations. The words in that set were useful for testing the hypotheses of interest, but they are too specific to the goals of the study and too few in number to constitute a general-use pool of well-characterized verbal stimuli.

Additionally, a focus on the survival-related characteristics of stimuli such as words can also be of value for exploring basic aspects of the representation of the concepts they denote in semantic memory (cf., Binder et al., 2016). Thus, a quantitative description that allows one to place a given concept along particular survival dimensions can contribute to a more thorough description of that concept’s set of componential features. Also, establishing the degree to which survival dimensions are related to other well-known, sensorimotor, semantic or emotional dimensions can be illuminating and reveal the ways in which interactions between these variables can modulate single-variable effects. Finally, the functional value of information about survival and how it relates to overall conceptual and lexical availability and accessibility could be a factor worth taking into account when addressing issues in lexical and semantic processing. As a matter of fact, dimensions related to survival, such as “danger” or “usefulness”, have already been demonstrated to affect performance in standard word recognition tasks such as naming and lexical decision tasks (Wurm, 2007, 2015; Wurm & Seaman, 2008). Expanding this type of study by incorporating a wider range of potentially relevant stimuli could lead to significant advances in the field.

The current study was designed to improve upon this stimulus-scarcity situation by collecting empirically derived normative data on the survival relevance of a relatively large set of concrete concepts, focusing on two specific dimensions. The selection of these two dimensions was based on data reported in two studies, one in English (Amsel, Urbach, & Kutas, 2012) and one in Spanish (Díez-Álamo, Díez, Alonso, Vargas, & Fernandez, 2018), that were aimed at collecting normative ratings for perceptual and motor dimensions of large sets of concrete concepts. In both cases, researchers analyzed the factorial structure of the sensorimotor properties of the rated words by means of a principal component analysis, finding that two factors had relatively high explanatory power in accounting for the variance. Interestingly, despite language and procedural differences, and despite only a partial overlapping between the referents of the two sets of words (approximately two thirds of the words referred to the same conceptual entities), the two factors proved to be identical in the two studies. More relevant for the present study, these factors were interpreted to be relevant to survival, and were labeled “avoiding death” and “locating nourishment” in the reports of the findings. The fact that two different survival-related factors emerged from the statistical computations suggests the possibility that survival relevance is a general property that can be separated into more concrete dimensions, such as those identified in the two mentioned studies. Therefore, and with the aim of obtaining an empirically based characterization of a set of verbal stimuli in terms of their particular survival value, a subjective rating procedure was implemented to collect independent data on each of the two selected dimensions, which in this study were labeled Avoiding Death (AD) and Obtaining Food (OF). Details of the design and methodology are described in what follows.

Method

Participants

A total of 300 Spanish-speaking undergraduate students pursuing Psychology or Speech Therapy degrees at the University of La Laguna (Spain) participated in the study, in exchange for course credit. Their ages ranged from 18 to 51 years (M = 20.6, SD = 3.23), and 84% were female.

Stimuli and materials

The set of 750 Spanish concrete words previously characterized in terms of perceptual and motor attributes by Díez-Álamo et al. (2018) conformed the pool of words to be normed in the present study (see Appendix 1 for an alphabetical list of the Spanish words and their English translation). The complete pool of words was partitioned into five sets to make the rating task more manageable, and each set was administered to 60 participants (30 per survival dimension). One of the sets consisted of 96 words that were initially rated due to their relevance for an ongoing memory study (not reported here). The 654 remaining words were randomly distributed into two sets of 164 words and two sets of 163 words. For each of these five word sets, three different word orders were randomly prepared and printed on multiple-page booklets. The first page in the booklet contained the rating instructions that asked participants to use a scale, from 1 to 5, to indicate their responses to the words in the following pages. The remaining pages included the list of words, organized into a single column, and each word with a 1-to-5 scale preprinted to its right.

Procedure

All the words were rated in two survival dimensions: avoiding death (AD) and obtaining food (OF), and data were collected in group sessions of between 10 and 20 participants that lasted approximately 20 minutes. Booklets of the two dimensions under study were randomly distributed to participants across sessions, with specific rating instructions printed on the first page of the booklets (the instructions are provided in Appendix 2). For participants in the AD condition, the instructions asked them to imagine being abandoned in an unknown place and attempting to survive by finding objects that could be of help in avoiding death by protecting from predators and other kinds of harm. For participants in the OF condition, the instructions also asked participants to imagine being abandoned in an unknown place, but this time they were asked to imagine attempting to survive by finding items of nourishment that could be drunk or eaten. All participants were then asked to consider each of the words in the booklet, following the preprinted order, and to use a pen to mark a position in the five-point scale on the right side of the word, indicating its relevance for the particular survival scenario that they were asked to imagine. Following Nairne et al. (2008), a value of 1 corresponded to “irrelevant”, and a value of 5 corresponded to “very relevant”, in the dimension being rated.

Results and discussion

Survival values for the 750 rated words in the two dimensions (AD and OF) were obtained by averaging the ratings provided by the participants (between 25 and 30 per word, with 30 independent ratings for 88% of the words). Potentially spurious order effects on the ratings were examined by determining the consistency of ratings to a given word across the three versions of the booklet in which it appeared, and ruled out on the basis of intraclass correlation coefficients [ICC (2,k)], which were calculated separately for the two dimensions. The results showed almost identical ICCs for all the rated word sets (see Table 1), with values ranging from 0.92 to 0.95 for the five booklets used in the AD rating task, and between .94 and .97 for the five booklets in the OF rating task. Split-half reliability was very high for both AD (Guttman split-half coefficient = .94) and OF ratings (Guttman split-half coefficient = .96).

Table 1 Intraclass correlation coefficients, by booklet, in each of the survival dimensions
Full size table

The mean survival value for the words rated in the AD dimension was 2.38 (SD = .91) in the five-point scale, with values ranging from 1.0 to 4.83. The mean survival value for the words rated in the OF dimension was also 2.38 (SD = 1.04) in the five-point scale, with values ranging from 1.0 to 4.90. The distribution of scores did not conform to a normal distribution in either of the two rated dimensions (Shapiro–Wilk, p < .001), being in both cases moderately positive skewed and, as shown in Fig. 1, with a higher probability density for lower ratings. The variables AD and OF were found to be positively and strongly correlated (r = .75; p < .001), reflecting the fact that many words had very similar values in the two dimensions, even though they were rated by different participants. With the data at hand, it is not possible to further investigate the nature or the number of potential relations underlying the correlation. But a reasonable interpretation would be that obtaining food can be taken as a particular case of a more general dimension such as avoiding death.

Fig. 1
figure 1

Violin plots showing the distribution and the probability density of rating scores for the normed set of 750 words in the survival dimensions of avoiding death (AD) and obtaining food (OF)

Full size image

A spreadsheet document provided with this report as supplementary material (SurvivalValue750words.xls) presents the relevant descriptive data for all the rated stimuli. In that document, words are listed in Spanish in alphabetical order in the first column, and columns to the right of each item present, in this order, the most common English translation of the word, its average rating in the AD dimension, the corresponding standard deviation, minimum and maximum average values, and the number of individual valid ratings per word in that dimension. Next, values for average rating, standard deviation, minimum, maximum, and number of contributing valid observations are provided for the OF dimension.

The first approach for exploring the nature of the survival ratings was to enter the survival scores in the two normed dimensions into correlational analyses along with a set of indexes of different types, obtained from a variety of recent normative studies in Spanish: psycholinguistic (number of letters, number of syllables, orthographic neighbors, written frequency, oral frequency, lexical decision, naming, familiarity, concreteness, imageability, subjective age of acquisition), perceptual (color vividness, visual motion, sound intensity, smell intensity, taste pleasantness), actional (graspability, body–object interaction), and emotional (valence, arousal, happiness, disgust, anger, fear, sadness). Table 2 shows the correlations with all those potentially relevant variables as well as the sources for the various indexes included in the correlational analyses.

Table 2 Correlations between the AD and OF survival dimensions and relevant variables
Full size table

Overall, significant but low correlations were found between survival values and some objective word descriptors, indicating that words that were more related to survival tended to be shorter in length and to have more orthographic neighbors. However, survival values were not significantly related to either the written or the oral frequency of words. Significant correlations emerged when a set of subjectively obtained word ratings were considered, showing that words that were related to survival in the two dimensions tended to be more familiar, more concrete, more imageable, and also more likely to be acquired early in the course of first-language acquisition.

Correlations with perception, action, and emotion variables were also observed. The AD dimension showed several significant but small correlations. Concepts that were considered more relevant for avoiding death in the imagined scenario were those that tended to have higher ratings in experiential dimensions, such as the likelihood of producing pain (r = .12), smell intensity (r = .26), graspability (r = .16), probability of body–object interaction (r = .26), and disgust (r = .16), while being less associated with feelings of sadness (r = −.13). The OF dimension demonstrated a somewhat different pattern of correlations, showing that concepts that are considered more relevant in the finding nourishment scenario tend to be perceived with more color vividness (r = .16), smell intensity (r = .52), taste pleasantness (r =.42), and visual motion (r = .15), and tend to be more graspable (r = .17), more likely to interact with the body (r = .23), and associated with higher disgust (r = .24) and less anger (r =−.14) and sadness (r = −.22).

To explore the nature of survival ratings in greater depth, two hierarchical regression analyses were conducted, with AD and OF ratings as the dependent variables, and a set of psycholinguistic, perceptual, action, and emotion variables as predictors, entered in successive steps. Table 3 shows the results of both analyses. The lexical and semantic predictor variables only accounted for 12% of the variance in AD and OF, with the entering of perception variables in the model producing a greater increment in the explained variance (ΔR2 =.22 for AD and ΔR2 =.47 for OF). Action variables only showed a significant but weak R2 increment for OF (ΔR2 =.03). Finally, emotion variables produced significant increments for both AD (ΔR2 =.09) and OF (ΔR2 =.05).

Table 3 Proportion of variation (adjusted R2) in AD and OF ratings explained by several predictor variables (hierarchical linear regression analyses)
Full size table

Overall, these results point to the importance of factors related to perception (e.g., smell, taste or pain), action (e.g., body–object interaction and graspability), and emotions (e.g., disgust, sadness), to explain the survival relevance ratings in both AD and OF scenarios. These results are in line with some findings in the field of memory that suggest that the differences in recall found in the survival processing literature may be related to a factor that can be regarded as “perceived threat to survival”. As Olds, Lanska, and Westerman (2013) point out, from this view, “survival processing might require participants to consider behaviors associated with avoiding bodily harm or sickness for self-preservation” (p. 34). Our results show that concepts that are considered relevant in the two survival scenarios are also associated with properties that could be related to avoiding both bodily harm (e.g., higher body–object interaction or graspability could be associated with the possibility of taking an object apart) and sickness (e.g., concepts more likely to produce pain, intense smell or taste, or provoking more disgust would be better for survival).

Finally, two three-step hierarchical linear regression analyses were performed to explore the power of survival values in predicting reaction times in lexical decision (Aguasvivas et al., 2018) and naming tasks (Davies et al., 2013). In step 1, several variables related to the lexical–word–form level were entered: number of syllables, orthographic neighborhood size and frequency, both written (Duchon, Perea, Sebastián-Gallés, Martí, & Carreiras, 2013) and oral (Alonso, Fernandez, & Díez, 2011). In step 2, two measures thought to be related to semantic processing were included, namely imageability (from Duchon et al., 2013) and age of acquisition (from Alonso, Díez, & Fernandez, 2016; and from Alonso, Fernandez, & Díez, 2015). Finally, in the last step, the survival ratings obtained in the present study were entered as predictors. As shown in Table 4, the amount of change in the variance predicted by survival ratings was low and nonsignificant both for lexical decision times and for naming times. In the light of these analyses, then, survival-related words do not have any significant predictive validity for response times reflecting lexical accessibility.

Table 4 Proportion of variation (adjusted R2) in lexical decision and naming response times explained by lexical and semantic variables, with AD or OF entered in the last step (hierarchical linear regression analyses)
Full size table

Conclusion

In this study, a relatively large set of words in Spanish were rated on two particular survival dimensions by Spanish-speaking young adults. As a result, quantitative estimations of the extent to which each of those words are related to the dimensions of Avoiding Death (AD) and Obtaining Food (OF) are now available to the research community. As described above, the procedure led to the collection of normative values that were very consistent across participants and highly reliable estimators for the two dimensions of interest. Although with the limitation of including only concrete nouns, the collected norms are likely to be particularly useful because of the rich characterization of the words in many other features and attributes, as quantitative indicators for their frequency of use, age of acquisition, category membership, affective value, bodily interaction, sensory experience, etc., are currently available. The data set thus has the potential to be a powerful tool for researchers conducting empirical studies with verbal stimuli in Spanish, both when controlling for potential stimulus-driven contaminating effects on their results and when their interest might be a purposeful manipulation of stimuli in investigating survival-relevant issues in cognition and, particularly, in memory.

It is a fact that the most frequently used empirical paradigm, the survival processing scenario (Nairne et al., 2007), has provided a wide range of interesting results. However, it is possible that the type of questions that can be addressed and the type of data that can be gathered can be substantially expanded by establishing research lines focused on the survival attributes inherent in the processed stimuli. As an example, studies analyzing a survival-related attribute such as animacy have contributed to a better understanding of important issues, for example, the status of survival value in relation to other semantic properties such as imagery (Bonin, Gelin, Laroche, Méot, & Bugaiska, 2015) or arousal (Popp & Serra, 2018), the relevance of survival value in processes such as categorization (Radanovic, Westbury, & Milin, 2016), or the neural correlates of the survival advantage in episodic memory (Xiao, Dong, Chen, & Xue, 2016). It is therefore foreseeable that other and more specific characterizations of survival-related properties in stimuli, as in the norms presented here, can lead to progress along similar lines of inquiry and open the way along which new lines can be explored. Explanatory accounts have made progress in the cognitive neuroscience of memory by testing hypotheses with empirical procedures aimed at obtaining convergent evidence, thus broadening the scope of reliable effects that can be brought to shed light on the validity of arguments. One illustrative example of such an approach in recent memory research is the study of the—also adaptation-related—phenomenon of using memory of the past to construct visions of the future (Schacter et al., 2012). Here, data from behavioral experiments (e.g., Ditta & Storm, 2016), neuroimaging techniques (e.g., Gilmore, Nelson, & McDermott, 2014), and neuropsychological samples (e.g., Hassabis, Kumaran, Vann, & Maguire, 2007) have been successfully combined to advance the identification of mechanisms and functions of episodic future thinking (Schacter, Benoit, & Szpunar, 2017). The availability of well-described materials, such as the set characterized for two survival dimensions in the present study, can be reasonably expected to be of value in setting a multi-approach research agenda in the realm of adaptive memory studies.

The database also has the potential for use in research programs focusing on more applied issues. For example, availability of stimuli of the kind presented here could be instrumental in further testing empirical hypotheses related to the general claim that, because our cognitive system is particularly primed to process survival information, the use of survival-relevant content could facilitate the early acquisition of basic concepts and skills in particular domains (e.g., the early learning of vocabulary in a foreign language; see Nairne, 2016; and VanArsdall, Nairne, Pandeirada, & Cogdill, 2015). Another field in which the database could be of relevance is in the scientific study of pathologies such as anxiety disorders, where findings are not totally consistent on whether patients have a tendency to process information about the world with a bias towards threat and danger (Mitte, 2008); here, having stimuli normed for a survival dimension, like death avoidance, could facilitate the development and use of empirical procedures aimed at better identifying risky personal profiles, and also conditions under which the malfunctioning of cognitive processes and representations could lead to pathological manifestations. Finally, and going a step beyond pure psychological interests, knowledge about how well-characterized survival-related information is perceived, acquired, retained, and used in human social environments could be useful in demonstrating the ways in which central aspects of human evolution can inform our understanding of social cognition (Schaller, Park, & Kenrick, 2006).

In conclusion, and from a more general and theoretical perspective, the characterization of the concepts in terms of the two dimensions of survival, avoiding death and obtaining food, can be seen as a step forward in the context of recent attempts to gather behavioral data on which to build empirically and brain-based componential semantic representations of concepts, taking into account relevant constituent features, their weight and combinatorial dynamics within a given concept, and their representational systems at a neural level (e.g., Binder et al., 2016). While most feature-based approaches typically incorporate associative properties, affective states, and sensorimotor attributes (e.g., Barsalou, 2012; Cree & McRae, 2003; Steyvers & Tenenbaum, 2005), the potential contribution of survival relevance as a defining conceptual feature has not been specifically targeted by semantic models or approaches. Whether it would eventually prove to be a conceptual aspect worth being taken into account can be considered as an issue for future investigation.