Emotional and temporal order effects – a comparison between word-cued and important autobiographical memories recall orders

ABSTRACT

The retrieval mechanisms associated with the recall of autobiographical memories (AMs) may differ according to the cueing method used to elicit AMs. In the present study, we provide a systematic comparison between word-cued and important AMs in terms of two recall order effects, namely a temporal and an emotional order effect. The sample comprised 104 adults aged between 59 and 90 years. On two measurement occasions, participants recalled up to 20 word-cued and 15 important AMs. For each memory, participants provided their age at the time when the event occurred (age-at-event) and rated its emotionality. Order effects were analysed based on multivariate multilevel model with autoregressive effects for emotionality and age-at-event, respectively. Whereas word-cued AMs showed a stronger emotional order effect, important AMs were recalled temporarily ordered. Individuals differed in the extent to which they ordered AMs along a temporal or emotional dimension. These differences could partly be explained by personality traits. For instance, higher scores on neuroticism were associated with a stronger emotional order effect in both memory types. Findings are discussed in terms of the retrieval mechanisms that are involved in the recall of word-cued and important memory recall and how they may be affected by personality.

KEYWORDS:

• Autobiographical memory
• retrieval mechanism
• recall order
• personality traits
• emotionality

In autobiographical memory research, two prominent types of memory elicitation methods are commonly used, namely the cue word method and the important memory method. The cue word method was developed by Crovitz and Schiffman (1974) as a modification of a procedure first described by Galton (1879). Here, participants are asked to retrieve autobiographical memories (AMs) that come to their minds when prompted with particular cue words, most often nouns (e.g., letter), provided by the experimenter (e.g., Haque & Conway, 2001; Rubin & Schulkind, 1997). Given that in everyday life, AMs can be activated by a variety of cues such as music (e.g., Jakubowski & Ghosh, 2021), odours (e.g., Chu & Downes, 2000), or pictures (El Haj et al., 2020), the cue word method is assumed to represent – at least to some extent – a naturalistic retrieval of AMs. Therefore, the method is considered to result in a putatively unbiased sample of recalled autobiographical events (Koppel & Berntsen, 2015). In the important memory method, participants are asked to recall events from their past that they consider personally important (e.g., Glück & Bluck, 2007; Rubin & Schulkind, 1997). Accordingly, memories elicited through the important memory method differ from more mundane (word-cued) personal experiences because they are more meaningful, more strongly related to personal identity (Glück & Bluck, 2007), and have a greater goal-relevance (Thomsen et al., 2012). Moreover, important memories form one’s life story (Köber & Habermas, 2017) and hence, are crucial for providing a sense of identity.

From these definitions, one might infer that the two methods involve different retrieval mechanisms (Koppel & Berntsen, 2019). There is some evidence showing that important AMs are typically retrieved temporarily forward ordered (e.g., Fromholt et al., 1995; Fromholt & Larsen, 1991), whereas the retrieval of word-cued AMs has been found to be associated with an emotional order effect, in the sense, that the emotionality of a successfully retrieved memory may function as a cue for subsequently retrieved AMs (e.g., Philippe et al., 2009, 2011). Hence, the important memory method may involve more strategic retrieval mechanisms (i.e., telling a temporarily coherent story about one’s life), whereas the cue word method encompasses more associative processes (i.e., AMs activate each other based on shared emotional aspects). However, a systematic comparison across the two methods regarding the retrieval mechanism each method triggers has not been done so far, and studies comparing important AMs and AMs retrieved in response to cue words (word-cued AMs) almost always focused on a comparison of the temporal memory distribution (e.g., Koppel & Berntsen, 2015, 2016; Rubin & Schulkind, 1997). Therefore, the present study adds to the literature by comparing two retrieval mechanisms across different types of AMs by systematically investigating the temporal and emotional order effect in word-cued and important AMs within subjects. Moreover, we extend previous research by introducing predictor variables that may explain individual differences within each retrieval mechanism.

Retrieval mechanisms of important and word-cued AMs

The autobiographical memory system is suggested to comprise four temporally defined levels of autobiographical knowledge layers of different specificity (Bluck & Habermas, 2000; Conway et al., 2019; Conway & Pleydell-Pearce, 2000). The lowest level occupies specific events, which are detailed records of experiences. The next level comprises general events referring to more fine-graded time periods. General events themselves are associated with lifetime periods defining overarching time spans within one’s life. Finally, the most abstract level of autobiographical knowledge is the life story represented as a schema, thereby contextualising specific memories within the whole life. This life story schema is a global, highly organised and generally stable representation of the life and is characterised by some sort of temporal chronology, given that it displays the temporal order of experiences that mimics the natural flow of time. Importantly, Bluck and Habermas (2000) assume the life story schema to be used as a basic template when important AMs has to be identified because this requires adopting a life perspective in comparing the memories within the context of one’s whole life and selecting the events according to their personal significance (Bluck & Habermas, 2000). Therefore, the important memory method is assumed to involve generative, top-down, and strategic processes (e.g., Bluck & Habermas, 2000; Koppel & Berntsen, 2015). Support for the assumption that the recall of important AMs produces a strategic, top-down, life story schema-based recall process can be found in findings of a chronological recall in important AMs (Fromholt & Larsen, 1991; Nusser et al., 2022; Nusser, Wolf, et al., 2023). For instance, Fromholt and Larsen (1991) investigated the temporal relation between successive important AMs and found that participants engaged more frequently in forward recall than backward recall. Nusser et al. (2022; Nusser, Zimprich, et al., 2023) extended this research by assessing the temporal order of the whole recall process instead of pairs of AMs. Their results indicated that participants started their recall with an important memory from an early age and moved forward in time. Moreover, individuals who recalled memories in a chronological order were also more likely to include memories from different life periods, thus covering a broader age span (Nusser et al., 2022). Doing so allows participants to tell a coherent story that describes personal continuity and development across the lifespan in a meaningful way, as well as explain changes in identity (Habermas & de Silveira, 2008; Köber et al., 2015). This seems especially important for older adults (Nusser, Wolf, et al., 2023), which might be ascribed to a higher motivation for stringing the AMs together into a meaningful story and establishing a sense of coherence across life (i.e., autobiographical reasoning; Habermas, 2011). Note that this refers to the recall of a longer sequence of important AMs (i.e., the life story). When it comes to single AMs, however, research has shown that older adults may be less (temporally) coherent (e.g., Sheldon et al., 2024 but see Siedlecki et al., 2015; Vannucci et al., 2021).

Taken together, the finding of a chronological recall order of important AMs provides some evidence for a strategic, top-down search based on the life story schema in which individuals adopt a lifetime perspective by recalling the most important events of their life in a temporally forward order. Even so, (older) adults are likely to order their important autobiographical memories along a temporal dimension, they might vary in the extent to which they temporally order their important AMs at recall (see also Nusser et al., 2022; Nusser, Zimprich, et al., 2023).

With respect to the cue word method, the precise nature of the retrieval mechanism is not fully understood (Koppel & Berntsen, 2019). A cue word can serve as a bottom-up stimulus triggering direct access to the specific event level (Conway, 2005; Conway & Loveday, 2010), a process referring to an associative, non-strategic, and direct bottom-up retrieval. This type of retrieval is more common in response to concrete cues that are high in imageability such as nouns (Uzer et al., 2012; Williams et al., 2006). Abstract cues and cues low in imageability such as emotional words, by contrast, are related to more generative search processes (e.g., Addis et al., 2012) starting at more abstract levels in the autobiographical memory system (Williams et al., 2006), although not necessarily at the highest level in the memory hierarchy (i.e., the life story schema). The information on these abstract levels are then used for further elaborative and iterative searches for more specific event-information (i.e., an AM). Notably, generative retrieval can also involves spreading activation across the memory network in an attempt to retrieve a memory that aligned with current goals (Harris et al., 2015). That is, memory cues activate each other until a cue is activated that successfully lead to the retrieval of the search-for memory from the autobiographical knowledge base (Conway, 2005). This suggests that even within generative retrieval, there exist random components characterised by automatic, associative, and non-strategic elements.

Support for the assumption of non-directed processes during the retrieval of word-cued AMs is apparent in the finding of word-cued AMs activating each other during the retrieval. This activation is based on shared characteristics such as time of the event (e.g., Brown & Schopflocher, 1998; Nusser & Zimprich, 2021), themes, or activities (e.g., Mace & Hall, 2018). Moreover, more abstract features of AMs may also play a role. For instance, Philippe et al. (2009, 2011) provided evidence that emotion (i.e., the emotionality of an AM) has the potential to act as an organising and binding principle during the recall process (see also Zimprich & Nusser, 2023). The authors asked participants to report a memory in response to an emotional cue. Afterward, participants recalled related AMs to the initial memory. Positive correlations were found between the positive (negative) valence of the initial memory and the positive (negative) valence of the linked memories, suggesting spreading activation within an emotional network that is activated during recall. A more recent study by Nusser and Zimprich (2021) extended this line of research by simultaneously examining both emotion and time as potential associative principles. Specifically, they investigated whether sequentially retrieved AMs are each associated with their respective predecessors based on the emotionality of the memory and the age at which the event took place. Results showed that individuals exhibited a “spill-over” effect (i.e., order effect) of both the emotionality as well as the age at the time of the current memory into subsequently reported AMs. Notably, the emotional order effect was more pronounced than the temporal order effect, indicating strong emotional links along which activation spreads during retrieval of word-cued AMs (emotional network model, Philippe et al., 2009). The fact that word-cued AMs are associated with each other although their retrieval process is “interrupted” by cue words strengthens the assumption that (also) associative processes come into play when the cue word method is used as memory elicitation method. It demonstrates that the memory system can not only be activated vertically (i.e., moving from abstract information to specific information and vice versa) (Sotgiu, 2021), but also horizontally (i.e., among content whose degree of abstraction is similar) (see also Mace & Clevinger, 2019).

Individual differences in retrieval mechanisms of word-cued and important AMs

Despite the growing number of literature on autobiographical memory, individual differences in the retrieval processes involved in memory recall have rarely been examined. However, there is first evidence demonstrating that individuals differ with respect to retrieval mechanisms for both the memory retrieval of word-cued AMs as well as of important AMs (e.g., Nusser et al., 2022; Zimprich & Nusser, 2023). For instance, Zimprich and Nusser (2023) showed that individuals differed in the extent to which word-cued AMs were ordered along an emotional dimension. In another line of research (Nusser et al., 2022; Nusser, Wolf, et al., 2023), the authors found that individuals also varied in the extent to which they ordered important AMs chronologically, with some individuals not doing it at all or to a limited extent. Following this line of research, the present study also examines individual differences in the retrieval processes associated with the recall of word-cued and important AMs in more detail by considering possible explanatory variables.

Potential candidate variables concerning individual differences in retrieval mechanisms may be personality traits. With respect to the emotional order, personality traits linked to general emotion processing, such as extraversion and neuroticism (Rubin et al., 2008), might have explanatory power. Extraversion is associated with positive affect and a greater orientation to positive and rewarding experiences (e.g., Morrone et al., 2000) and an overall positive mindset. Aligned with that, previous studies have demonstrated that extraversion correlated with the recollection of positive AMs (e.g., Denkova et al., 2012), providing evidence for positive AMs offering higher accessibility and pre-activation in individuals scoring high on extraversion. By contrast, neuroticism was found to be related to emotionally more negative AMs (Denkova et al., 2012; Rasmussen & Berntsen, 2010, Study 1) and a tendency to ruminate about negative aspects (Teasdale & Green, 2004). Therefore, neuroticism has been suggested to involve negative affective memory networks (Robinson et al., 2007; Rusting, 1998), which was supported by priming studies (e.g., Robinson et al., 2007). On this basis, the personality traits extraversion and neuroticism may be related to a stronger tendency to order word-cued AMs along an emotional dimension.

Another personality trait that is strongly related to autobiographical remembering is openness to new experiences. In general, openness is linked to dimensions of complexity (McCrae & Costa, 1997) and to interest in exploring and being aware of new (inner) experiences. With respect to AMs, openness is related to the self-rated overall usage of AMs in the sense that higher scores on openness are related to a more frequent use of AMs for different purposes (e.g., Cappeliez & O'Rourke, 2002; Rasmussen & Berntsen, 2010). More importantly, previous research demonstrated a consistent relationship between openness to experience and a higher coherence of single AMs (Rasmussen & Berntsen, 2010), as well as with a higher structural complexity of narratives (McAdams et al., 2004). Finally, individuals open to experiences are more likely to perceive growth in interpreting the past (Bauer et al., 2005; see also Pals, 2006). Predictions regarding the effect of openness to experience on order effects during the recall of AMs seem less straightforward. Given the association between openness to experiences and the way individuals tell their personal past, it could be speculated that individuals high in openness may retrieve their AMs in such ways that help them to increase narrative coherence in their life story, for instance, by retrieving their AMs in a temporal order.

The present study

The present study aimed to compare the retrieval processes involved in word-cued versus important AMs. To that end, we investigated the emotional and the temporal recall order of AMs across both methods of elicitation (cue word vs. important) in older adults with the former one representing associative processes and the latter one depicting strategic processes. The focus on an older age sample offers the advantage that older adults have access to a much more varied pool of life events kept in autobiographical memory.

As a consequence, their AMs differ more regarding many attributes. For instance, memories show a larger range of participants’ ages when the event happened as well as differ more regarding their emotionality (Mickley & Kensinger, 2009; St. Jacques & Levine, 2007), which facilitates the investigation of temporal and emotional order effects. Note that one line of research revealed that older adults tend to recall event memories with lower specificity, resulting in fewer event-specific details being remembered (e.g., Addis et al., 2008; Levine et al., 2002). This age-related memory decline was especially shown for episodic details (e.g., time and place) of mundane AMs, such as word-cued AMs. Episodic details of important AMs, however, might be better preserved due to the higher personal relevance and the strong relation to personal identity (e.g., Glück & Bluck, 2007), the greater goal-relevance, and the higher frequency of rehearsal (Thomsen et al., 2012). Hence, we would expect that the recall of word-cued AMs might be less temporarily ordered, but instead show an emotional order effect. By contrast, research on the recall of important AMs almost exclusively focussed on a temporal order (i.e., chronology) as one retrieval mechanism. Hence, we expected important AMs to show a strong temporal order effect, whereas predictions regarding an emotional order effect in important AMs seem less straightforward. As such, the present study aimed to replicate and extend previous research by systematically investigating both retrieval mechanisms, i.e., the temporal and emotional order effect, across the two differently cued memory types. Note that we do not expect important memory recall to involve solely strategic processes and word-cued memory recall to rely exclusively on associative processes. Instead, we assume that, in comparison, important memory recall involves more strategic processes than word-cued memory recall, as demonstrated by a stronger temporal order effect. Conversely, we expect word-cued memory recall to be characterised by a greater involvement of associative processes compared to important memory recall, as evidenced by a more pronounced emotional order effect.

Moreover, we used a within-subject design; that is, older adults retrieved both important and word-cued AMs. This has two advantages: First, errors associated with individual differences are reduced. Second, individual tendencies, in the sense of higher emotionality or age-at-event ratings, as well as of emotional or temporal order tendencies, across types of memories can be examined. Following previous research, we assumed the retrieval of word-cued AMs to be ordered along an emotional dimension. By contrast, important AMs are expected to be retrieved in a (chronological) temporal order. By applying a multivariate multilevel model with autoregressive effects, we took the dependency of the data due to repeated measurements into account. Moreover, the statistical approach allowed us to investigate and compare the order effects across types of memories on the within- and the between-person level. Based on previous research, we expected differences in the extent to which individuals order their AMs along an emotional and temporal dimension (e.g., Nusser, Wolf, et al., 2023; Nusser & Zimprich, 2021). Such individual differences may be explained by genuine individual difference variables such as personality traits. Whereas neuroticism and extraversion may affect the emotional order of AMs, predictions regarding openness are less straightforward and examined in a more explorative fashion.

Methods

Participants

Participants were recruited through ads in local newspapers in Southern Germany, flyers and word of mouth from 2020 to 2021. Study participation required a working knowledge of German language. After excluding 17 persons who only reported important or cue word AMs, the sample comprised 104 (59.6% female) older adults aged between 59 and 90 years (M = 69.34, SD = 6.54).¹ More than half of the participants reported being married (57.7%), 19.2% were divorced, 15.4% were widowed, and 7.7% reported being single (see Table 1). Furthermore, 83.7% of the participants had one or more children. 63.5% of the participants lived with their partner or another family member, one person reported living in a residential care home, and 35.6% lived alone. Regarding educational background, 38.5% of the participants reported having graduated from university, and 15.4% of participants graduated from high school (equals approximately 13 years of formal education). 30.8% reported having Realschulabschluss, and 15.4% finished Haupt-/Volksschule, which corresponds to 10 and 9 years of formal education.

Table 1. Sample characteristics.

	M (SD)	Range
Age	69.34 (6.54)	59–90
Extraversion	2.20 (0.58)	1.00–3.45
Openness	2.59 (2.59)	1.42–3.92
Neuroticism	1.36 (0.68)	0.17–3.33
	n	%
Gender (female)	62	59.6
Marital status
Married	60	57.7
Divorced	20	19.2
Widowed	16	15.4
Single	8	7.7
Children^a	87	83.7
Living conditions
With partner or family member	58	63.5
Alone	37	35.6
Residential care home	1	1
Educational background
University	40	38.5
High School	16	15.4
Realschule^b	32	30.8
Volks-/Hauptschule^c	16	15.4

^a Reflects the number and percentage of participants answering “yes” to this question.

^b corresponds to 10 years of formal education.

^c corresponds to 9 years of formal education.

Procedure and measures

Data come from a comprehensive study focusing on different types of long-term memories in older adulthood.² The study was conducted in accordance with the Declaration of Helsinki and approved by the ethical committee of [anonymized for review]. In the following, we describe the measures and the procedure that are relevant to the present research. Word-cued and important AMs were assessed in an interview-like setting by research assistants, who had been trained regarding the study procedure and the tasks used in the present study. The two memory tasks were performed on different measurement occasions with at least four weeks in between in order to avoid carry-over effects. The order in which participants performed these tasks (either starting with important or word-cued AMs) was randomised across participants.

Cue word autobiographical memory task

The task included 20 cue words. Each cue word was presented on a small card. The order in which cue word cards were presented was randomised for each participant anew. After being presented with a cue word, participants were asked to briefly report the first personal experience evoked by that cue word. The interviewer wrote down each memory description in keywords on separate documentation sheets. Participants were instructed that the personal events did not have to be extraordinary but specific and distinct. To assure that the AMs reported would fulfil these criteria, participants were provided with an example describing possible experiences that could be associated with the cue word “fire.” Moreover, if participants described a general or extended event, the interviewer asked them to focus on a specific episode within that event. If a participant was unable to retrieve a specific event memory in response to a cue word, the interviewer went on to the next cue word card. There was no strict time limit, but participants were asked to end with the description of a memory after approximately two minutes.

The 20 cue words were drawn from the Berlin Affective Word List (BAWL-R; Võ et al., 2009). Cue words were selected based on high imageability and emotional neutrality. On a scale ranging from not at all imageable (1) to completely imageable (7), the 20 cue words had a mean of 6.37, indicating high imageability. On a scale ranging from completely negative (−3) to completely positive (3), the selected cue words had a mean of 0.45, indicating emotional neutrality. For a list of all 20 words see Appendix Table A1.

Important autobiographical memory task

Participants were asked to recall fifteen important events from their lives and briefly describe each event. The interviewer wrote down a short description of each event on separate documentation sheets. Participants were instructed that the AMs did not have to be extraordinary but refer to a specific and distinct event from their past that they consider personally important. If participants described a general or extended event, the interviewer asked them to choose a specific episode from that event. After having described an important AM, the interviewer asked the participant to describe another important life event. This procedure was repeated until a participant had described a total of fifteen important AMs or could not come up with another important event. There was no strict time limit, but participants were asked to end with the description of a memory after approximately two minutes.

Memory ratings

After the AMs (i.e., cue word AMs, and important AMs, respectively) had been described, documentation sheets were randomised (i.e., shuffled). Participants were then prompted with each of their AMs and asked to report their age at the time of the event as well as to judge the emotionality of the events on a 7-point Likert-type scale ranging from very negative (1) to very positive (7). The procedure and the rating questions were identically for both memory tasks.

Personality traits

Personality traits were measured using three subscales of the German Revised NEO-Personality Inventory (NEO-FFI; Borkenau & Ostendorf, 1993; Costa & McCrae, 1992), which capture individuals’ levels of extraversion (12 items), neuroticism (12 items), and openness (12 items). Questions assess the extent to which people agree or disagree with statements that describe them. Responses were made on a scale ranging from disagree strongly (0) to agree strongly (4). Items were averaged to produce subscales indicative of three personality traits, namely extraversion (Cronbach’s alpha = .80), neuroticism (Cronbach’s alpha. = .85), and openness (Cronbach’s alpha. = .60).

Statistical approach

In order to model a temporal and emotional order effect in word-cued and important AMs, a multivariate model with first-order autoregressive effects was used. In an autoregressive model, the measurement of an outcome variable in a longitudinal sequence (here: order of output) is described as a function of previous outcomes of the same variable (see, e.g., Diggle et al., 2002). Since we analyse two autoregressive effects (i.e., emotional and temporal order effect) in both word-cued and important AMs, we have four separate models in which the previous measurement of age-at-event and emotionality, respectively, enters the model as a first-order autoregressive effect (AR-effect). Notably, in the present study, the focus is on autoregressive effects within persons. For this reason, we centred the lag variables around their respective latent cluster (i.e., person) mean (cf. Gistelinck et al., 2021). Moreover, because the present data exhibit a two-level structure (Level 1: important and word-cued autobiographical memories, Level 2: individuals), we extended the multivariate model with first-order autoregressive effects by including random effects for the intercepts and the slopes (autoregressive effects) to take the hierarchical nature of the data into account (see Hedeker, 2008; Hedeker & Gibbons, 2006) and to capture individual differences in the average of age-at-event and emotionality, respectively, as well as individual differences in the strength of the two order effects. The first-order autoregressive model then becomes a multilevel autoregressive model (or mixed-effects autoregressive model, Figure 1). More specifically, let $y_{\mathrm{in}}$ denote the valence (or age) of the memory reported by participant i at output position n, ${\mu }_{\xi }$ the fixed intercept, ${\xi }_i$ the random intercept effect of participant i, ${\mu }_{\rho }$ the fixed autoregressive effect, ${\rho }_i$ the random autoregressive effect of participant i, and ${ϵ}_{\mathrm{in}}$ the residual (or error), we have that $y_{\mathrm{in}}={\mu }_{\xi }+{\xi }_i+({\mu }_{\rho }+{\rho }_i){\eta }_{i(n-1)}+{ϵ}_{\mathrm{in}}(n>1).$Here, ${\eta }_{i(n-1)}$ is the valence (or age) of the memory reported by participant i at output position $n-1$ that has been mean-centred around the (latent) intercept of participant i, that is, ${\eta }_{i(n-1)}=y_{i(n-1)}-({\mu }_{\xi }+{\xi }_i).$Doing so (instead of using the observed participant-specific mean) helps to avoid Nickell’s bias (Nickell, 1981), an underestimation of the fixed effect estimate of the autoregressive parameter. Figure 1 depicts the model as a structural equation model, implying that many software packages can be used to estimate it. A more detailed description of the model is given in Zimprich and Nusser (2023).

Figure 1. Model with random intercept (ξ) and random first-order autoregressive (ρ) effects. Note that the observed indicators (Y₁ to Y_n) are centred around the latent group mean (i.e., the random intercept).

Results

Descriptive statistics

Instead of the possible 20 (word-cued AMs) × 104 (participants) + 15 (important AMs) x 104 (participants) = 3640 AMs, there were only 3505 AMs because some participants were unable to generate a memory to some of the cue words or were unable to report 15 important AMs. Furthermore, the exclusion of 57 AMs that had missing values on the age-at-event or emotionality variables led to a total sample of 3448 AMs, of which 1949 AMs were word-cued AMs and 1499 were important AMs. On average, participants recalled 18.74 word-cued (range word-cued AMs: 12–20) and 14.40 important AMs, respectively (range important AMs: 9–15). Word-cued and important AMs differed in their mean age-at-event, with word-cued AMs having a higher mean age-at-event (M = 31.08, SD = 21.87) than important AMs (M = 29.5, SD = 18.78; t(3446) = 2.22, p < .05, Cohen’s d = .08). There were no significant differences in terms of emotionality (M_{word-cued AMs} = 4.68, SD = 2.15; M_{important AMs} = 4.61, SD = 2.42).

In terms of the personality variables, participants reported moderate levels of extraversion and openness, and lower levels of neuroticism (see Table 1). Extraversion was positively related to openness (r = .268, p < .001) and negatively associated with neuroticism (r = −.372, p < .05). Openness and neuroticism were not significantly related (r = −.067, p = .502).

Main analyses

Autoregressive effect of emotionality

As can be seen in Table 2,³ the intercept of emotionality was 4.68 for word-cued and 4.54 for important AMs, which indicates that, overall, participants rated their AMs as being slightly more on the positive side. Furthermore, the random intercept variance of both types of memories was significantly different from zero, implying that individuals differed in their average ratings of emotionality in word-cued and important AMs.

Table 2. Parameter estimates of mixed-effects autoregressive models.

	Word-cued AMs (N = 1749)			Important AMs (N = 1383)
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
Emotionality of AMs as DV
Fixed effects
Intercept (${\mu }_{\xi }$)	4.682*	4.676*	4.674*	4.540*	4.541*	4.543*
AR-effect (${\mu }_{\rho }$)	0.161*	0.159*	0.155*	0.031	0.030	0.030
Age^b		−0.023*	−0.022*		0.008	0.005
Sex^b		−0.087	−0.081		−0.322*	−0.248^†
Extraversion^b			0.229*			−0.009
Extraversion → Slope^b			−0.030			0.011
Neuroticism^b			−0.150*			−0.284*
Neuroticism → Slope^b			0.095*			0.081*
Openness^b			0.099			0.057
Openness → Slope^b			−0.047			−0.032
Random Effects
Intercept variance $\left({\sigma }_{\xi }^2\right)$	0.154*	0.148*	0.144*	0.234*	0.225*	0.212*
Slope variance $\left({\sigma }_{\rho }^2\right)$	0.021*	0.020*	0.019*	0.024*	0.021^†	0.020^†
$\mathrm{co}{r}_{\eta \left(\rho \tilde{y}\right)}$	−0.101	−0.117	−0.129	0.143	0.154	0.166
Error variance $\left({\sigma }_{ϵ}^2\right)$	4.225*	4.216*	4.212*	5.423*	5.422*	5.417*
Model fit
−2LL	7543	7536	7509	6330	6324	6301
AIC	7555	7552	7537	6342	6340	6329
Age at Event as DV
Fixed effects
Intercept (${\mu }_{\xi }$)	31.001*	31.077*	31.059*	34.607*	34.611*	34.442*
AR-effect (${\rho }_{\tilde{x}}$)	0.047^†	0.048^†	0.050^†	0.624*	0.624*	0.618*
Age^b		0.167	0.229		0.001	−0.019
Sex^b		1.272	0.354		−0.045	0.424
Extraversion^b			0.536			−0.136
Extraversion → Slope^b			−0.075^†			0.070^†
Neuroticism^b			2.845*			−1.980*
Neuroticism → Slope^b			0.019			−0.028
Openness^b			1.588			0.492
Openness → Slope^b			−0.091^†			0.002
Random effects
Intercept variance $\left({\sigma }_{\xi }^2\right)$	87.319*	86.012*	82.608*	51.525*	51.217*	41.442*
Slope variance $\left({\sigma }_{\rho \tilde{x}}^2\right)$	0.005	0.006	0.001	0.071*	0.070*	0.061*
$\mathrm{co}{r}_{\xi \left(\rho \tilde{x}\right)}$	−0.128	−0.229	−0.352	0.852*	0.871*	0.890*
Error variance $\left({\sigma }_{\delta }^2\right)$	383.273*	383.287*	383.432*	194.877*	194.862*	192.643*
Model fit
−2LL	15532	15530	15503	11313	11313	11292
AIC	15544	15546	15531	11325	11329	11320

Note: DV = dependent variable; b = variable was centred across persons (grand-mean centred); sex was dummy-coded with 0 (female) and 1 (male); −2LL = −2 × Log-Likelihood (smaller is better); AIC = Akaike’s Information Criterion (smaller is better).

*p < .05; ^†p < .10.

The AR-effect of emotionality was only significant for word-cued memories. The AR- effect of emotionality describes that a more positive memory was followed by another more positive AM. Note that because emotionality was centred within participants, the autoregressive effect implies that a memory with an emotionality rating above an individual’s mean tended to be followed by another memory above an individual’s mean emotionality rating, while a memory with an emotionality rating below an individual’s mean tended to be followed by another memory below average emotionality. This is illustrated in Figure 2.

Figure 2. Illustration of the AR-effect of emotionality across output positions for word-cued and important autobiographical memories.

Note: The dotted line depicts the predicted emotionality of word-cued AMs; the solid line depicts the predicted emotionality of important AMs. Horizontal lines depict intercept estimates.

As shown in Table 2, the random variance estimates of the AR-effects of emotionality were significant for both word-cued and important AMs, implying that individuals differed in the extent to which they emotionally ordered their AMs.

In Model 2, age of participant and sex were included as control variables. As Table 2 shows, age had a negative and significant effect on the emotionality intercept of word-cued AMs. This implies, that with increasing age more negative word-cued AMs were reported. By contrast, age had no effect on the emotionality intercept of important AMs. Sex had no effect on emotionality in word-cued AMs but a negative one in important AMs. That is, men reported more negative important AMs than women. Compared to Model 1, model fit was improved in both types of memory as indexed by the −2ℓℓ and AIC fit statistics. This improvement in fit was significant (word-cued AMs: Δ−2ℓℓ = 7, Δ df = 2, p < .05; important AMs: Δ−2ℓℓ = 6, Δ df = 2, p < .05).

In Model 3, we examined whether extraversion, neuroticism and openness to experiences had an effect on the strength of the AR-effect of emotionality. To do so, effects of personality traits on the (latent) AR-effect of emotionality were included. The according parameter estimated can be found in Table 2. Regarding extraversion, there was only a significant effect on the average emotionality of word-cued AMs implying that individuals with higher extraversion ratings also reported more positive word-cued AMs. With respect to neuroticism, we found a negative effect on the AR-effect of emotionality of both word-cued and important AMs, which indicates that more neurotic individuals tend to recall more negative AMs. Moreover, individuals with higher ratings on neuroticism had a stronger AR-effect of emotionality in word-cued AMs as indicated by the significant effect of neuroticism on the AR-effect. In important AMs, the effect of neuroticism on the AR-effect of emotionality became significant as well. This implies that although the AR-effect of emotionality is insignificant, it gets stronger (and potentially significant) for those individuals who score high on neuroticism. No effects were found for openness to new experiences. For both types of memory, Model 3 showed a considerable and significant improvement in fit compared to Model 2 (word-cued AMs: Δ−2ℓℓ = 27, Δ df = 6, p < .05; important AMs: Δ−2ℓℓ = 23, Δ df = 6, p < .05).

Autoregressive effect of age-at-event

Regarding age-at-event, on average, participants were 31 years at the time of the word-cued AMs and 34.61 years at the time of the important AMs (see Table 2). The random intercept variances were significant, which indicates reliable individual differences in the age-at-event ratings for both word-cued and important AMs. The AR-effect of age-at-event became significant for important AMs only. Thus, an important memory with an age-at-event rating above (or below) an individual’s mean tended to be followed by another important memory above (or below) an individual’s average age-at-event rating. Figure 3 aims to illustrate the AR-effect of age-at-event across output positions. Again, the predicted values of age-at-event for word-cued and important AMs are shown as dotted and solid lines, respectively. As can be seen, in important AMs the age-at-event increases across output position describing a series of adjacent important AMs with similar age-at-event values (i.e., above or below an individual’s average). That is, participants do not only cluster important AMs that occurred close in time. They also order them chronologically in the sense that they start with an event stemming from an early age and then move forward in time. By contrast, there isn’t such a strong trend in word-cued AMs.

Figure 3. Illustration of the AR-effect of age-at-event across output positions for word-cued and important autobiographical memories.

Note: The dotted line depicts the predicted age-at-event of word-cued AMs; the solid line depicts the predicted age-at-event of important AMs. Horizontal lines depict intercept estimates.

As can be seen in Table 2, The random variance estimate of the AR-effect of age-at-event was significant for important AMs. Hence, there were individual differences in the extent to which participants ordered their important AMs according to a temporal dimension.

In Model 2, control variables were added but neither age nor sex had an effect on the age-at-event intercept in word-cued AMs as in important AMs. The AR-effects of age-at-event did not change after including age and sex. Accordingly, model fit of Model 2 did not improved significantly compared to Model 1 (word-cued AMs: Δ−2ℓℓ = 2, Δ df = 2, p > .05; important AMs: Δ−2ℓℓ = 0, Δ df = 2, p > .05). Thus, Model 2 did not capture the data better than Model 1. Notwithstanding, in order to control for both age and sex, we included these two demographic predictor variables in all subsequent models.

Again, in Model 3 participants’ ratings on extraversion, neuroticism and openness to experiences scales and their effects on the AR-effect were included as potential predictor variables to explain individual differences in the age-at-event variable and its AR-effect. In terms of extraversion, there was an effect of extraversion on the AR-effect of age-at-event in both, word-cued and important AMs, which was significant only at the p < .10 level. Regarding neuroticism, we found an effect on the age-at-event intercept which was positive in word-cued AMs but negative in important AMs. This implies that individuals with higher ratings on neuroticism tend to report word-cued AMs from an older age and important AMs from a younger age (as compared to individuals who score lower on neuroticism). There was no effect of neuroticism on the AR-effect of age-at-event. Openness to new experiences had a negative effect on the AR-effect of age-at-event in word-cued AMs which, however, was only significant at the p < .10 level and led only to a diminution of the marginally significant AR-effect. As Table 2 shows, including the personality traits as predictors led to a substantial and significant improvement in model fit compared to Model 2 (word-cued AMs: Δ−2ℓℓ = 27, Δ df = 6, p < .05; important AMs: Δ−2ℓℓ = 21, Δ df = 6, p < .05).

Discussion

Previous studies suggest that different retrieval mechanisms are associated with the retrieval of word-cued and important AMs. Whereas important AMs are typically ordered temporarily (e.g., Fromholt & Larsen, 1991; Nusser et al., 2022) – especially in older age (e.g., Nusser, Wolf, et al., 2023), word-cued AMs not only show a temporal order but an even more pronounced emotional order effect (e.g., Nusser & Zimprich, 2021). The present study aimed to add to the existing literature by investigating both recall order effects across differently cued AMs within individuals. As such, the present study provides a more thorough examination of the retrieval mechanisms involved in the recall of word-cued and important AMs within a sample of older adults. Moreover, we explored whether individual differences in both order effects could be explained by personality traits such as extraversion, neuroticism, and openness.

Emotional and temporal order effects

As expected, we found an emotional order effect in word-cued but not in important AMs. This finding might indicate that the retrieval of word-cued AMs involves more associative processes compared to the retrieval of important AMs. To be more specific, in word-cued AMs individuals may guide their recall in the sense that the retrieved memory is associated with the cue word. However, also random components may come into play that are based on spreading activation within the pre-existing knowledge structure of autobiographical memory and that are not directed by the individuals themselves. These activations can be of different types. Our findings of an emotional order effect demonstrated that word-cued AMs can activate each other based on emotional associations. Moreover, we also found a temporal order effect, but only tendentiously. From this finding, one could infer that word-cued AMs are more strongly activated along an emotional rather than a temporal line during recall and that emotional aspects serve as effective cues for subsequently retrieved AMs (Schulkind & Woldorf, 2005). This resembles Unsworth et al. (2014) notion of a search framework in long-term memory that is constituted by a directed and a random component. The former refers to strategic processes that can be controlled by the individual. The latter describes the probabilistic nature of the search process, implying that information is activated and spread automatically through the information generated by the directed strategy (Shiffrin, 1970; Shiffrin & Atkinson, 1969; Unsworth et al., 2014).

A theoretical rationale for the finding that the emotional order effect exceeded the temporal order effect might be found in the context retrieval model (Cohen & Kahana, 2022; Talmi et al., 2019), proposing that emotional items activate each other because they share the same context (e.g., internal features such as feelings and physiological changes). Hence, physiological reactions that are elicited while remembering an emotional event might additionally serve as a retrieval cue for the next memory. Accordingly, the emotional network of AMs would not only be activated through the recall of an emotional memory but also through current physiological arousal. Put differently, physiological reactions might reinforce the activation of the emotional network of AMs and thus increase the probability of recalling an emotional memory. Notably, the finding of associative processes in the sense that AMs can activate each other via emotional (and temporal) associations does not imply that the retrieval of AMs in response to cue words does not include generative memory retrieval (e.g., Wank et al., 2021). Instead, we argue here that the retrieval mechanisms associated with the recall of AMs in response to cue words – compared to those involved in the recall of important AMs – are less strategy-based. Rather, they appear to be based on automatic processes such as spreading activation within the underlying autobiographical memory structure. This may be especially the case in studies that use cue words high in imageability and instruct participants to recall specific event memories (e.g., Williams et al., 2006; see also Conway & Jobson, 2012) as it was done in the present study.

With respect to the recall of important AMs, we found a temporal order effect. That is, important AMs were ordered along a temporal dimension in the sense that the age at the time when a remembered event took place predicted the age-at-event of the subsequently recalled memory. From that, one could conclude that individuals start their recall with any important memory and then move forward to another important memory that had happened close in time. However, our findings revealed that the recall of important AMs is not simply structured along a temporal dimension in the sense that events that happened together are recalled together; rather, it appears as if important AMs are recalled chronologically by starting with a memory that happened early in life and ending with a memory from the more recent past (see Figure 2), thus covering a person’s (entire) life span. This aligns with previous research demonstrating a chronological order effect in the recall of important AMs (e.g., Nusser, Wolf, et al., 2023).

Given the absence of any emotional order effect, we posit that, in contrast to the retrieval of word-cued AMs, the retrieval of important AMs is subject to lesser influence from automatic search processes. Instead, it is likely guided by a more strategic search process (Koppel & Berntsen, 2015) and a deliberate, intentional ordering of memory recall (Bluck & Habermas, 2000). When asked to recall important AMs, the recalled information is required to be interpreted and reflected upon within the context of a larger view of one’s life in order to identify those experiences that are perceived as most relevant to one’s life and self. One way of doing so is to establish a broad temporal context of important AMs and to report them in a chronological temporal order (e.g., Bluck & Habermas, 2000; Habermas & de Silveira, 2008). Recalling the important events of one’s life in a temporal order can be assumed to be a “diachronic” autobiographical reasoning process that reflects the continuity of the self over time (Bluck & Habermas, 2000). This might be especially the case in samples of older adults (Nusser, Wolf, et al., 2023). As postulated by Erikson (1950, 1982), older adults are confronted with the psychosocial theme of ego-integrity, described as “the acceptance of one’s one and only life cycle as something that had to be” (Erikson, 1950, p. 268) and as “a sense of coherence and wholeness” (Erikson, 1982, p. 65). That is, the development of ego-integrity involves an attempt to make sense of one’s own past and to gain a sense of meaning from life, which can be done by reviewing the past in a meaningful and coherent way, for instance by remembering important life-story events in a forward temporal order. Hence, from a theoretical perspective, it seems functional to recall important AMs in a temporal or chronological manner in order to create coherence and meaning in life (in older age). One has to keep in mind, though, that the present findings provide only indirect support for the assumption that the recall of important AMs involves strategic recall mechanisms, and further research is needed to understand the underlying (motivational) factors that guide the chronological order effect found in the recall of important AMs.

Effects of personality on memory retrieval mechanisms

Apart from examining whether different cueing methods (i.e., word cued vs. important AMs) result in different retrieval mechanisms (i.e., emotional vs. temporal order effect), we explored whether individual differences in the emotional and temporal order effect could be explained by personality. As expected, we found that personality traits linked to general emotion processing (i.e., extraversion and neuroticism) are associated with emotional characteristics of AMs. For instance, individuals scoring higher on extraversion reported emotionally more positive word-cued AMs, whereas individuals scoring higher on neuroticism reported emotionally more negative word-cued and important AMs. Hence, these two personality traits are associated with the overall emotional tone, respectively the individual mean emotionality of the AMs retrieved – a finding that aligns with previous research (e.g., Denkova et al., 2012; Rasmussen & Berntsen, 2010).

In terms of their potential effects on the emotional output order of AMs, and as expected, we found neuroticism to be related to a person’s tendency to order their AMs along an emotional dimension. Specifically, with increasing levels of neuroticism, the emotional order effect becomes stronger. Notably, this held not only for word-cued AMs, for which an emotional order effect has been shown before (e.g., Philippe et al., 2011; Zimprich & Nusser, 2023), but also for the recall of important AMs. Hence, for people high in neuroticism, the emotionality of a memory may function as an organising and binding principle during the recall of AMs – irrespective of the retrieval process (associative, non-strategic vs. more strategic, directed retrieval). Considering that neuroticism is not only associated with a stronger emotional order of AMs during retrieval, but also with the recall of emotionally more negative memories, our findings (further) support the assumption that the memory network of individuals scoring high on neuroticism is biased toward negative affective information (Robinson et al., 2007; Rusting, 1998). Against our expectations, extraversion was unrelated to the emotional order effect, implying that the degree to which AMs are ordered emotionally during recall is independent from an individual’s level of extraversion. We grounded our argument on research showing an increased orientation towards positive and rewarding experiences in people scoring high on extraversion (e.g., Morrone et al., 2000). However, extraversion is not only characterised by the experience of positive emotions, but also on the general tendency to seek stimulation from the outside world and to get attention from other people. As such, further research is needed to examine whether and how extraverted individuals order their (word-cued) AMs during recall.

Regarding the temporal order effect, personality traits seem less influential. In terms of openness to new experience, this finding somewhat contradicts our expectations. Previous research has shown that openness is positively related to the structural complexity of narratives (McAdams et al., 2004) and higher coherence of single memories (Rasmussen & Berntsen, 2010). Based on these findings, we considered the possibility that individuals high in openness may apply a stronger temporal order during memory retrieval – for instance, in order to tell a coherent and meaningful life story. The temporal order effect found for the recall of important AMs, however, was unrelated to openness. This implies that the extent to which important AMs are ordered along a temporal dimension is independent of individual levels of openness. Instead, our findings suggest that openness may affect the extent to which word-cued AMs are ordered temporarily, but opposite to the expected direction: The temporal order effect in word-cued AMs seems tendentiously stronger for participants scoring low on openness. Considering that the retrieval of AMs in response to cue words is assumed to primarily involve an associative, non-strategic search process (Koppel & Berntsen, 2015), age at the time when the event occurred may become an important cue for the subsequently recalled memory for those participants, who are less open. In contrast, individuals high in openness, which can be described as fantasy prone, creative, and independent-minded (Rasmussen & Berntsen, 2010), may make use of all kinds of associations between subsequently retrieved AMs and not show one particular output order. Note however, that the effect of openness was small and only marginally significant – as was the temporal order effect in word-cued AMs. As such, these findings need to be interpreted with caution.

Notably, the personality trait of extraversion had an effect on the temporal order effect in both word-cued and important AMs. In terms of word-cued AMs, this effect was negative implying that individuals scoring high on extraversion tended to recall word-cued AMs less temporarily ordered compared to individuals who score low on extraversion. Again, one could argue that the recall of word-cued AMs in extraverted individuals is driven by other associations among AMs and, therefore, is neither showing a temporal nor an emotional order effect. Given that extraverted individuals are socially oriented, word-cued AMs may be associated via shared social themes. Up to this point, this is only speculative in nature, but could be the target of future research. With respect to important AMs, we found extraversion to enhance the temporal order effect in the sense that individuals high in extraversion show a stronger tendency to retrieve their most important AMs in a temporal or chronological order. This finding could be explained by the functional use of AMs. Individuals high in extraversion tend to use their AMs more frequently to socially connect with others (e.g., Rasmussen & Berntsen, 2010). For instance, McLean and Pasupathi (2006) showed that extraverted people are more comfortable sharing their AMs with others and enjoy mutual reminiscence more than less extraverted people. Extraverted people also share their self-defining memories with more people. Frequently talking about self-defining, respectively the most important events experienced in life may provide individuals with more coherence and certainty in their life stories (cf. McLean & Pasupathi, 2006), thus explaining the stronger temporal order effect found for individuals high in extraversion.

Limitations and future directions

The present study investigated the retrieval mechanisms in older adults. Our focus on old age was motivated by the fact that older adults (compared to younger adults) have access to a much more varied pool of AMs. As a consequence, one may expect that, in principle, their AMs differ more regarding many attributes, e.g., emotionality and time. Nevertheless, the investigation of age differences in the recall order of word-cued and important AMs would contribute to a better understanding of the underlying mechanisms of the retrieval in autobiographical memory. With respect to the recall of important AMs, previous research found that the temporal order of important AMs differs as a function of age in the sense that older adults had a much more pronounced temporal order than younger adults (Nusser, Wolf, et al., 2023). Age differences in the recall order of word-cued AMs have not been studied yet. On the one hand, research indicates that there are no age effects in directly recalled memories, which are based on automatic memory recall and spreading activation processes (Markostamou et al., 2023, also see the influential functional account of memory and aging by Craik, 1986). On the other hand, according to the socioemotional selectivity theory (Carstensen et al., 1999), which states a relative preference for positive over negative information also in autobiographical memory (e.g., Kennedy et al., 2004; Ros & Latorre, 2010), one might deduce that the emotional recall order is more pronounced in older than younger adults.

Apart from age, there might be other person-related aspects that influence the recall order of word-cued or important AMs. The present study provides evidence for individual differences in the average emotionality ratings and age-at-event ratings, as well as in the emotional and temporal order effect. Moreover, some of these individual differences could be explained by personality traits. Nevertheless, the investigation of individual differences in the recall order of word-cued and important AMs might be of particular interest and value for future research. Future studies might extend the present study by including other individual differences measures such as dispositional self-foci (Teasdale & Green, 2004) or time perspective (Mello et al., 2009) in order to explain within-person tendencies in emotionality and age-at-event across word-cued and important memories as well as their potential association with order effects.

Additional explanatory variables for individual differences in both the emotional as well as the temporal order effect might involve basic cognitive abilities, which were not assessed in the present study. According to Unsworth et al. (2013), working memory control processes are of vital importance for a controlled search in long-term memory. They demonstrated that individuals with high working memory capacity are more strategic than individuals with low working memory capacity because they excel at dynamically utilising search strategies and generating, then elaborating on cues that specify the search set (Spillers & Unsworth, 2011; Unsworth & Engle, 2007). Similarly, individual differences in the emotional order effect and associative recall processes, respectively, may stem from a deterioration in associative memory, describing difficulties in forming and retrieving associations between elements of an event (e.g., Naveh-Benjamin, 2000; Old & Naveh-Benjamin, 2008). The extension of these findings to associations among distinct events requires further investigation. Nevertheless, less efficient binding may be associated with a lowered reliance on associative processes during memory recall. Future research might include measures of cognitive functioning (e.g., working memory capacity) potentially influencing individual differences in strategic and associative recall processes in general and, specifically, in the temporal and emotional order effects.

Conclusion

Taken together, by systematically investigating the recall order of word-cued AMs and important AMs within persons, the present study demonstrated that both cueing methods could be contrasted according to which retrieval mechanisms each method triggers (Koppel & Berntsen, 2019). As the emotional order effect in word-cued memories indicates, emotional aspects of a memory can activate other emotionally congruent memories. Hence, retrieval mechanisms in word-cued AMs are rather non-strategic and automatic. By contrast, the forward temporal order effect in the important memory recall suggests the involvement of strategic, narrative-based search processes, such as activating a life story schema. Finally, the present research points to individual differences in the extent to which individuals order their AMs along an emotional and temporal dimension (e.g., Nusser, Wolf, et al., 2023; Nusser & Zimprich, 2021) which could be partly explained by personality traits such as neuroticism.

Open practices statements

The data as well as the syntax that support the findings are available on request from the corresponding author. Data were analysed using SAS, Version 9.4 (SAS Institute Inc, 2015). This study’s design and its analysis were not pre-registered.

Acknowledgements

The order of authors is on a nominal scale; all authors have contributed equally.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Notes

1 As multilevel models are rather complex, conducting a power analysis for multilevel models (with autoregressive effects) is complicated and unambiguous conclusions are difficult to make in many cases (Kreft & de Leeuw, 1998). The sample size of the present study was guided by general suggestions for the use of multilevel models (e.g., Maas & Hox, 2005) as well as by our previous studies in which we investigated order effects in AMs (e.g., Zimprich & Nusser, 2023; Nusser & Zimprich, 2021).

2 Parts of the data (i.e., important AMs) have been published with respect to recall trajectories of important AMs (see Nusser et al., 2022) and developmental themes represented in important AMs (Nusser, Zimprich, et al., 2023).

3 For the following analysis, the AMs at the first output position were excluded as only those AMs that had at least one “predecessor” can be used in order to form a first-order AR-effect. Hence, the numbers of the AMs as well as the estimates for the mean age-at-event, emotionality respectively (i.e., intercept) displayed in Table 2 differ from those provided in the descriptive statistics.

Appendix

Table A1. List of cue words (translation in parentheses) and their ratings according to emotional valence and imageability as taken from the Berlin Affective Word List (Võ et al., 2009).

Word (noun)	Valence		Imageability
	Mean	SD	Mean	SD
Bank (bank)	−0.24	1.16	6.35	1.16
Boot (boat)	1.05	1.43	6.05	1.29
Brief (letter)	1.50	0.90	6.42	0.86
Fernsehen (television)	0.21	1.43	6.42	0.90
Fliege (fly)	−0.95	1.10	6.14	1.17
Frosch (frog)	−0.09	1.54	6.05	1.13
Kaffee (coffee)	1.40	0.97	6.78	0.44
Kastanie (chestnut)	0.91	1.23	6.05	1.21
Keller (basement)	0.00	0.82	6.33	0.71
Lineal (ruler)	0.30	0.48	6.44	1.13
Loch (hole)	−0.90	0.74	6.11	1.05
Löffel (spoon)	0.59	0.96	6.14	1.28
Lupe (magnifier)	0.00	0.00	6.56	0.53
Markt (market)	0.80	0.79	6.22	0.67
Regen (rain)	0.32	1.77	6.77	0.71
Stern (star)	1.00	0.94	6.67	0.50
Straße (street)	0.18	0.83	6.46	0.99
Wald (forest)	1.50	0.85	6.56	0.73
Zeitung (newspaper)	0.60	1.07	6.11	1.05
Zitrone (lemon)	0.80	0.79	6.78	0.44

Note: emotional valence was rated on a 7-point scale ranging from −3 (very negative) through 0 (neutral) to +3 (very positive). The 7-point imageability scale ranged from 1 (low imageability) to 7 (high imageability).