INFLUENCE OF EMOTIONAL VALENCE AND ... - Semantic Scholar
INFLUENCE OF EMOTIONAL VALENCE AND AROUSAL ON THE SPREAD OF ACTIVATION IN MEMORY Sandra JHEAN-LAROSE *, Nicolas LEVEAU ** and Guy DENHIERE ** * Université d'Orléans - Laboratoire EDA - Education Discours Apprentissage Université Paris Descartes - 45, Rue des Saints-Pères - 75006 Paris, France
** Équipe CHArt – Cognition Humaine et Artificielle 41 rue Gay Lussac – 75005 Paris, France Abstract: Controversy still persists on whether emotional valence and arousal influence cognitive activities. Our study sought to compare how these two factors foster the spread of activation within the semantic network. In a lexical decision task, prime words were varied depending on the valence (pleasant or unpleasant) or on the level of emotional arousal (high or low). Target words were carefully selected to avoid semantic priming effects, as well as to avoid arousing specific emotions (neutral). Three SOA durations (220ms, 420ms and 720ms) were applied across three independent groups. Results indicate that at 220ms, the effect of arousal is significantly higher than the effect of valence in facilitating spreading activation while at 420ms, the effect of valence is significantly higher than the effect of arousal in facilitating spreading activation. These findings suggest that affect is a sequential process involving the successive intervention of arousal and valence. Key words: Emotion, cognition, lexical decision task, semantic memory, associative network
*
[email protected]
**
[email protected]
**
[email protected]
1.
INTRODUCTION
Much of the research on the psychology of emotions has been concerned with the effect of valence, that is, on how pleasant or unpleasant emotional experiences impact behavior (Bodenhausen, Kramer, & Süsser, 1994 , Abele, Gendofla, & Petzold, 1998; Krauth-Gruber & Ric, 2000; Bestgen, 2002; Corson 2002a; White 2007). Recent models suggest that emotions are episodes of synchronized organismic changes (Scherer, 1984, 2005; Niedenthal, KrauthGruber & Ric, 2009) and cannot be strictly characterized as a hedonic value. There have been numerous attempts at modeling, which have led to the introduction of other dimensions in the characterization of emotions such as physiological activation or "arousal" (Russell, 1980; Larsen & Diener, 1992; Russell & Feldman Barrett, 1999; Russell, 2003), action tendencies (Frijda, 1986), or control and novelty of the stimulus (Fontaine, Scherer, Roesch, & Ellsworth, 2007). In their study, Fontaine et al. (2007) conclude that valence and arousal can account for 46.7% of the observed variance between two distinct emotions. However, these models focus primarily on steady–state emotions rather than on dynamic stimuli. For Scherer (1984), affective states are a response to endogenous and exogenous stimuli that individuals analyze sequentially and systematically. The earliest mode corresponds to the assessment of the novelty of the stimulation in relation to previous experiences, and then comes the assessment of the hedonic orientation of the stimulation. Affective priming (Fazio, Sanbonmatsu, Powell, & Kardes, 1986; Bower, 1991; Klauer, 1997), an experimental paradigm inspired by semantic priming (Neely, 1977), makes it possible to manipulate the automatic or strategic character of the cognitive processes implemented in a requested task. After presenting an emotionally-connoted stimulus (for example, the word “aggressive”), participants are requested to evaluate an emotionally connoted target (for example, the word “rose”). The evaluation of the target could include, for example, estimating its emotional valence or could be on whether it belongs to the French language or not. The evaluation task therefore requires one to be knowledgeable on the semantic properties of the word presented. The time interval between the presentation of the target and the presentation of the source determines the nature of the cognitive process in play. The main advantage of priming techniques in investigating emotional processes lies in the fact that they overcome the main criticisms against emotional evaluation techniques by making it possible to analyze the automatic and spontaneous character of participants’ responses, that is, by considering that affective judgment precedes the cognitive strategic judgment of a stimulus (Zajonc, 1980, 1984). Olofsson, Nordon, Seiqueira and Piloch (2008) have summarized ERP studies within the last 40 years. They highlighted that in these studies, valence exerts influence predominantly between 100 and 300 ms, whereas arousal exerts influence from 200 ms and later. However, if simultaneous effect of valence and arousal controlled stimulus characteristics have been studied, it is for 400-800 ms latency range only, and it revealed significant effect of the sole arousal characteristic. Hinojosa, Carretié, Méndez-Bértolo, Mígues and Pozo (2009) have used high- and low-arousing congruent, and high- and lowarousing incongruent positive pairs for a priming ERP experiment (SOA = 300 ms). Participants had to press a button as quickly and accurately as possible, to tell if the target name was arousing or relaxing. No effect of arousal was observed. However, in the difference of typical lexical decision task, instructions here involve strategic long latency process and might be not congruent to arousal assessment dynamics. Rossell and Nobre (2004) conducted an affective priming experiment using lexical decision task with neutral, happy, fearful and sad pairs. 200 ms, 700 ms and 950 ms SOA were used. For short or medium SOA (200ms and 700 ms), the authors observed significant difference between sad and fear (low- vs high-
arousal) when prime and target pairs were related; no significant difference was observed for unrelated prime-target pairs. For longer SOA (950 ms), the opposite pattern appeared: significant difference between sad and fear was only observed for unrelated pairs. The variation of the influence sad and fear stimulus within time could therefore be explained either by the fact that they refer to different emotion categories (discrete emotions), or by the fact that they refer to different arousal degrees (dimensional emotions). However, for longer SOA (950 ms), the authors did not observe significant result for semantically unrelated pairs suggesting the preponderant role of semantic vs emotional prime characteristics. From the cognitive processes point of view of emotions, Forgas (1995) proposes the AIM (“Affect Infusion Model”) model. Affect infusion can be defined as a process whereby affectively loaded information exerts an influence on the judgmental process, altering deliberations and outcomes. This model assumes that the nature and extent of the influence of affect on judgment is largely dependent on the type of process chosen by a judge. It distinguishes between judgments using processes sensitive to affective states from those using processes insensitive to the affective state. It identifies four types of processes: the first two are sensitive to affect infusion: direct access which involves reproducing a stored reaction, and motivated processing which implements predetermined patterns of information search and makes little use of innovative and constructive processes. The last two processes, heuristic and substantive, are insensitive to affect infusion and require a high degree of creativity. Consequently, affect infusion focuses on automated processes rather than on strategic, analytical or over-learned processes. Experimental research mainly based on mood induction has investigated the influence of affect on accessing information in memory (Bower, Monteiro, & Gilligan, 1978; Bower, 1981 ; White, 2007). Corson’s (2002b) observation that positive moods promote access to general knowledge can be explained by reduced spreading time between one conceptual node and another, that is, by the increase in permeability of the associative network. Hänze et Hesse (1993) highlighted this increase in semantic network permeability by using a lexical decision task (SOA of 200 ms) with participants induced into positive or neutral moods and with semantically associated pairs. Results indicated a significant decrease in response time for strongly associated pairs when participants were induced into positive moods, compared to when participants were induced into neutral moods. No significant differences were observed among participants in positive or neutral moods for weakly associated pairs. However, Isen and Daubman (1984) suggest that the increase in associative network permeability resulting from a positive mood leads to a richer and more complex cognitive context. The large quantity of conceptual nodes activated at any given moment decreases the available cognitive resources. According to these authors, this decrease offers a cognitive explanation of the preferential use of heuristic rather than strategic strategies when individuals are in a positive mood. It is for this reason that this rich cognitive context promotes the implementation of more flexible and creative processes that seek to reduce the arising cognitive load. Consequently, positive moods intervene across two successive periods: first they lead to a decrease in cognitive resources, and then to the compensatory implementation of categorization processes of the concepts activated. Moreover, if the associative network is defined by a semantic pre-activation preceding a positive mood, the cognitive context is enriched to a lesser extent by the increase in the spreading of semantic memory than if this activation had not been implemented. In this case, there will be no reduction in cognitive resources and a positive mood will shorten the time taken to access information in memory. On the contrary, without pre-activation, positive moods will increase the permeability of the associative network, and therefore lead to a richer cognitive context and to a decrease in cognitive resources. Subsequently, it will take longer to access information in memory.
Given that induced states (generally joy and sadness) in many of the studies comparing the influence of pleasant and unpleasant affective states differ in both valence and arousal, Corson (2006) used a lexical decision task to study how the modification of arousal influences cognitive processes independently from valence orientation. Joy (Pleasant / High Arousal), anger (Unpleasant / High Arousal), sadness (Unpleasant / Low Arousal) and relaxation (Pleasant / Low Arousal) induced moods were considered. Participants were asked to judge primes and targets; the next word was presented 100ms after the previous word had been judged (McNamara & Altarriba, 1988). Corson did not conclude in terms of discrete but in terms of dimensional emotion: He came to the conclusion that facilitated spreading within the highly associated semantic network occurred for high arousal moods whereas facilitated spreading was not observed in low arousal moods. Nevertheless, the dynamic aspect of activation and emotional valence on observed behavior has not been addressed. Below 300 ms, semantic priming is attributed to automatic processes (Posner & Snyder, 1975; Neely, 1977; Ratcliff & McKoon, 1981). While affective priming effects are observed below SOA 300 ms with lexical primes, they become more moderate as from 500 ms (for a review see Klauer, 1997), and disappear as from 1000 ms (Hermans, De Houwer , & Eelen, 1994). None of the researches mentioned above considered the effect of the emotional characteristics of representations in semantic memory. The main objective of our study is to highlight the temporal dynamics of the components of emotional valence and arousal on affect, as well as to determine when they are activated in semantic memory and for how long. Our main assumptions concern the effect of (i) Arousal, (ii) Valence and (iii) their respective interaction with SOA relative to response times in a lexical decision task. Our study sought to analyze how valence and emotional arousal of a lexical stimulus influence spreading activation in memory. To prevent from any semantic priming effect, we have chosen unrelated prime-target pairs. To prevent for affective priming effect, we have chosen neutral target (medium valence and arousal). The SOA durations were fixed at 220 ms, 420 ms and 720 ms, ranging from automatic to strategic processes. As regards emotional valence, Hänze and Hesse (1993) argue that positive valence leads to faster spreading of associative memory due to an increase in permeability of the associative network. Isen and Daubman (1984) highlight the reduction of available cognitive resource when participants are in positive mood, resource that should preferably be allocated to the high associated semantic sub-network. For unrelated positive emotional valence of prime will thus be accompanied by longer response times. As valence comes into play secondarily in the emotional process (Scherer, 1984), the modulation of the cognitive function by affect will therefore be, secondly, preferentially due to emotional valence. Positive emotional valence of the prime will be accompanied by an increase in lexical decision reaction time that will be longer at an SOA of 420ms than at an SOA of 220ms. However, as affective priming effects are more moderate as from SOA 500ms (Klauer, 1997), the increase in lexical decision time will be inferior at an SOA of 720ms than at an SOA of 420ms. As regards arousal, Corson’s (2006) findings reveal that high arousal leads to faster spreading of associative memory activation due to an increase in associative network permeability. For similar reasons as for emotional valence, an increase in arousal of the prime will therefore be accompanied by longer response times. However, according to Scherer (1984), affective states are a result of a sequential process in which the first two steps are assessing stimulus novelty, and then evaluating the unpleasantness or the pleasantness of the situation. The modulation of the cognitive function by the affective state will be therefore,
initially, preferentially due to arousal. High arousal of the prime will be accompanied by a more important increase in the lexical decision reaction time for a SOA of 220ms than for longer SOA. 2.
2.1
METHOD
PARTICIPANTS Eighty-four native French adults voluntarily participated in the study.
The datasets of two participants were excluded as average of their response times was more than three standard deviations from the mean (305 ms for the SOA = 220 ms group, 284 ms for the SOA = 420 ms group, 115 ms for the SOA = 720 ms group). Participants with an error rate superior to 25% were also excluded. There were 27 participants per group. The average education level of the participants was 3.4 years after the high school diploma for the 220ms group, 3.7 years for the 420ms group and 3.4 years for the 720ms group. There were 41% women and 59% men for the 220ms group and the 420ms group, and 70% women and 30% men for the 720ms group). 2.2
MATERIAL
The experimental material was established based on Leleu’s semantic atlas of emotional concepts (Leleu, 1987; Hogenraad & Bestgen, 1989) and on the LEXIQUE database (New, Pallier, Ferrand, & Matos, 2001). Leleu’s (1987) atlas includes 3000 words; the words arousal and emotional valence have been evaluated by 39 judges on average on a scale ranging from 1 to 7. For each word, an integer value of valence and arousal ranging from 10 to 70 is proposed. As this is an unpublished norm, we verified its relevance by comparing the data of similar words with Valemo’s norm (Syssau & Font, 2005). Results indicated a correlation of valence of .93 (p
** Équipe CHArt – Cognition Humaine et Artificielle 41 rue Gay Lussac – 75005 Paris, France Abstract: Controversy still persists on whether emotional valence and arousal influence cognitive activities. Our study sought to compare how these two factors foster the spread of activation within the semantic network. In a lexical decision task, prime words were varied depending on the valence (pleasant or unpleasant) or on the level of emotional arousal (high or low). Target words were carefully selected to avoid semantic priming effects, as well as to avoid arousing specific emotions (neutral). Three SOA durations (220ms, 420ms and 720ms) were applied across three independent groups. Results indicate that at 220ms, the effect of arousal is significantly higher than the effect of valence in facilitating spreading activation while at 420ms, the effect of valence is significantly higher than the effect of arousal in facilitating spreading activation. These findings suggest that affect is a sequential process involving the successive intervention of arousal and valence. Key words: Emotion, cognition, lexical decision task, semantic memory, associative network
*
[email protected]
**
[email protected]
**
[email protected]
1.
INTRODUCTION
Much of the research on the psychology of emotions has been concerned with the effect of valence, that is, on how pleasant or unpleasant emotional experiences impact behavior (Bodenhausen, Kramer, & Süsser, 1994 , Abele, Gendofla, & Petzold, 1998; Krauth-Gruber & Ric, 2000; Bestgen, 2002; Corson 2002a; White 2007). Recent models suggest that emotions are episodes of synchronized organismic changes (Scherer, 1984, 2005; Niedenthal, KrauthGruber & Ric, 2009) and cannot be strictly characterized as a hedonic value. There have been numerous attempts at modeling, which have led to the introduction of other dimensions in the characterization of emotions such as physiological activation or "arousal" (Russell, 1980; Larsen & Diener, 1992; Russell & Feldman Barrett, 1999; Russell, 2003), action tendencies (Frijda, 1986), or control and novelty of the stimulus (Fontaine, Scherer, Roesch, & Ellsworth, 2007). In their study, Fontaine et al. (2007) conclude that valence and arousal can account for 46.7% of the observed variance between two distinct emotions. However, these models focus primarily on steady–state emotions rather than on dynamic stimuli. For Scherer (1984), affective states are a response to endogenous and exogenous stimuli that individuals analyze sequentially and systematically. The earliest mode corresponds to the assessment of the novelty of the stimulation in relation to previous experiences, and then comes the assessment of the hedonic orientation of the stimulation. Affective priming (Fazio, Sanbonmatsu, Powell, & Kardes, 1986; Bower, 1991; Klauer, 1997), an experimental paradigm inspired by semantic priming (Neely, 1977), makes it possible to manipulate the automatic or strategic character of the cognitive processes implemented in a requested task. After presenting an emotionally-connoted stimulus (for example, the word “aggressive”), participants are requested to evaluate an emotionally connoted target (for example, the word “rose”). The evaluation of the target could include, for example, estimating its emotional valence or could be on whether it belongs to the French language or not. The evaluation task therefore requires one to be knowledgeable on the semantic properties of the word presented. The time interval between the presentation of the target and the presentation of the source determines the nature of the cognitive process in play. The main advantage of priming techniques in investigating emotional processes lies in the fact that they overcome the main criticisms against emotional evaluation techniques by making it possible to analyze the automatic and spontaneous character of participants’ responses, that is, by considering that affective judgment precedes the cognitive strategic judgment of a stimulus (Zajonc, 1980, 1984). Olofsson, Nordon, Seiqueira and Piloch (2008) have summarized ERP studies within the last 40 years. They highlighted that in these studies, valence exerts influence predominantly between 100 and 300 ms, whereas arousal exerts influence from 200 ms and later. However, if simultaneous effect of valence and arousal controlled stimulus characteristics have been studied, it is for 400-800 ms latency range only, and it revealed significant effect of the sole arousal characteristic. Hinojosa, Carretié, Méndez-Bértolo, Mígues and Pozo (2009) have used high- and low-arousing congruent, and high- and lowarousing incongruent positive pairs for a priming ERP experiment (SOA = 300 ms). Participants had to press a button as quickly and accurately as possible, to tell if the target name was arousing or relaxing. No effect of arousal was observed. However, in the difference of typical lexical decision task, instructions here involve strategic long latency process and might be not congruent to arousal assessment dynamics. Rossell and Nobre (2004) conducted an affective priming experiment using lexical decision task with neutral, happy, fearful and sad pairs. 200 ms, 700 ms and 950 ms SOA were used. For short or medium SOA (200ms and 700 ms), the authors observed significant difference between sad and fear (low- vs high-
arousal) when prime and target pairs were related; no significant difference was observed for unrelated prime-target pairs. For longer SOA (950 ms), the opposite pattern appeared: significant difference between sad and fear was only observed for unrelated pairs. The variation of the influence sad and fear stimulus within time could therefore be explained either by the fact that they refer to different emotion categories (discrete emotions), or by the fact that they refer to different arousal degrees (dimensional emotions). However, for longer SOA (950 ms), the authors did not observe significant result for semantically unrelated pairs suggesting the preponderant role of semantic vs emotional prime characteristics. From the cognitive processes point of view of emotions, Forgas (1995) proposes the AIM (“Affect Infusion Model”) model. Affect infusion can be defined as a process whereby affectively loaded information exerts an influence on the judgmental process, altering deliberations and outcomes. This model assumes that the nature and extent of the influence of affect on judgment is largely dependent on the type of process chosen by a judge. It distinguishes between judgments using processes sensitive to affective states from those using processes insensitive to the affective state. It identifies four types of processes: the first two are sensitive to affect infusion: direct access which involves reproducing a stored reaction, and motivated processing which implements predetermined patterns of information search and makes little use of innovative and constructive processes. The last two processes, heuristic and substantive, are insensitive to affect infusion and require a high degree of creativity. Consequently, affect infusion focuses on automated processes rather than on strategic, analytical or over-learned processes. Experimental research mainly based on mood induction has investigated the influence of affect on accessing information in memory (Bower, Monteiro, & Gilligan, 1978; Bower, 1981 ; White, 2007). Corson’s (2002b) observation that positive moods promote access to general knowledge can be explained by reduced spreading time between one conceptual node and another, that is, by the increase in permeability of the associative network. Hänze et Hesse (1993) highlighted this increase in semantic network permeability by using a lexical decision task (SOA of 200 ms) with participants induced into positive or neutral moods and with semantically associated pairs. Results indicated a significant decrease in response time for strongly associated pairs when participants were induced into positive moods, compared to when participants were induced into neutral moods. No significant differences were observed among participants in positive or neutral moods for weakly associated pairs. However, Isen and Daubman (1984) suggest that the increase in associative network permeability resulting from a positive mood leads to a richer and more complex cognitive context. The large quantity of conceptual nodes activated at any given moment decreases the available cognitive resources. According to these authors, this decrease offers a cognitive explanation of the preferential use of heuristic rather than strategic strategies when individuals are in a positive mood. It is for this reason that this rich cognitive context promotes the implementation of more flexible and creative processes that seek to reduce the arising cognitive load. Consequently, positive moods intervene across two successive periods: first they lead to a decrease in cognitive resources, and then to the compensatory implementation of categorization processes of the concepts activated. Moreover, if the associative network is defined by a semantic pre-activation preceding a positive mood, the cognitive context is enriched to a lesser extent by the increase in the spreading of semantic memory than if this activation had not been implemented. In this case, there will be no reduction in cognitive resources and a positive mood will shorten the time taken to access information in memory. On the contrary, without pre-activation, positive moods will increase the permeability of the associative network, and therefore lead to a richer cognitive context and to a decrease in cognitive resources. Subsequently, it will take longer to access information in memory.
Given that induced states (generally joy and sadness) in many of the studies comparing the influence of pleasant and unpleasant affective states differ in both valence and arousal, Corson (2006) used a lexical decision task to study how the modification of arousal influences cognitive processes independently from valence orientation. Joy (Pleasant / High Arousal), anger (Unpleasant / High Arousal), sadness (Unpleasant / Low Arousal) and relaxation (Pleasant / Low Arousal) induced moods were considered. Participants were asked to judge primes and targets; the next word was presented 100ms after the previous word had been judged (McNamara & Altarriba, 1988). Corson did not conclude in terms of discrete but in terms of dimensional emotion: He came to the conclusion that facilitated spreading within the highly associated semantic network occurred for high arousal moods whereas facilitated spreading was not observed in low arousal moods. Nevertheless, the dynamic aspect of activation and emotional valence on observed behavior has not been addressed. Below 300 ms, semantic priming is attributed to automatic processes (Posner & Snyder, 1975; Neely, 1977; Ratcliff & McKoon, 1981). While affective priming effects are observed below SOA 300 ms with lexical primes, they become more moderate as from 500 ms (for a review see Klauer, 1997), and disappear as from 1000 ms (Hermans, De Houwer , & Eelen, 1994). None of the researches mentioned above considered the effect of the emotional characteristics of representations in semantic memory. The main objective of our study is to highlight the temporal dynamics of the components of emotional valence and arousal on affect, as well as to determine when they are activated in semantic memory and for how long. Our main assumptions concern the effect of (i) Arousal, (ii) Valence and (iii) their respective interaction with SOA relative to response times in a lexical decision task. Our study sought to analyze how valence and emotional arousal of a lexical stimulus influence spreading activation in memory. To prevent from any semantic priming effect, we have chosen unrelated prime-target pairs. To prevent for affective priming effect, we have chosen neutral target (medium valence and arousal). The SOA durations were fixed at 220 ms, 420 ms and 720 ms, ranging from automatic to strategic processes. As regards emotional valence, Hänze and Hesse (1993) argue that positive valence leads to faster spreading of associative memory due to an increase in permeability of the associative network. Isen and Daubman (1984) highlight the reduction of available cognitive resource when participants are in positive mood, resource that should preferably be allocated to the high associated semantic sub-network. For unrelated positive emotional valence of prime will thus be accompanied by longer response times. As valence comes into play secondarily in the emotional process (Scherer, 1984), the modulation of the cognitive function by affect will therefore be, secondly, preferentially due to emotional valence. Positive emotional valence of the prime will be accompanied by an increase in lexical decision reaction time that will be longer at an SOA of 420ms than at an SOA of 220ms. However, as affective priming effects are more moderate as from SOA 500ms (Klauer, 1997), the increase in lexical decision time will be inferior at an SOA of 720ms than at an SOA of 420ms. As regards arousal, Corson’s (2006) findings reveal that high arousal leads to faster spreading of associative memory activation due to an increase in associative network permeability. For similar reasons as for emotional valence, an increase in arousal of the prime will therefore be accompanied by longer response times. However, according to Scherer (1984), affective states are a result of a sequential process in which the first two steps are assessing stimulus novelty, and then evaluating the unpleasantness or the pleasantness of the situation. The modulation of the cognitive function by the affective state will be therefore,
initially, preferentially due to arousal. High arousal of the prime will be accompanied by a more important increase in the lexical decision reaction time for a SOA of 220ms than for longer SOA. 2.
2.1
METHOD
PARTICIPANTS Eighty-four native French adults voluntarily participated in the study.
The datasets of two participants were excluded as average of their response times was more than three standard deviations from the mean (305 ms for the SOA = 220 ms group, 284 ms for the SOA = 420 ms group, 115 ms for the SOA = 720 ms group). Participants with an error rate superior to 25% were also excluded. There were 27 participants per group. The average education level of the participants was 3.4 years after the high school diploma for the 220ms group, 3.7 years for the 420ms group and 3.4 years for the 720ms group. There were 41% women and 59% men for the 220ms group and the 420ms group, and 70% women and 30% men for the 720ms group). 2.2
MATERIAL
The experimental material was established based on Leleu’s semantic atlas of emotional concepts (Leleu, 1987; Hogenraad & Bestgen, 1989) and on the LEXIQUE database (New, Pallier, Ferrand, & Matos, 2001). Leleu’s (1987) atlas includes 3000 words; the words arousal and emotional valence have been evaluated by 39 judges on average on a scale ranging from 1 to 7. For each word, an integer value of valence and arousal ranging from 10 to 70 is proposed. As this is an unpublished norm, we verified its relevance by comparing the data of similar words with Valemo’s norm (Syssau & Font, 2005). Results indicated a correlation of valence of .93 (p
