strategies: an MRI study
Esra C.S. de Groot
University of Amsterdam
Esra C.S. de Groot, Department of Developmental Psychology, University of Amsterdam.
Correspondence concerning this article should be addressed to Esra C.S. de Groot, Department of Developmental Psychology, University of Amsterdam. E-mail:
2 Abstract
Previous research has found consistent individual differences in social learning strategies and social information use. However, the underlying causes and mechanisms of the individual differences in social learning strategies are still poorly researched. Therefore, this research aims to investigate the relation between individual differences in social learning strategies, the environment of
development, and brain structure and connectivity. Participants finished a questionnaire (Risky Families Questionnaire) to measure the environment of development, a computer task (BEAST) to measure social information use, and brain structure and connectivity was measured with an MRI-scan. The neural focus of this research lays on the anterior cingulate cortex (ACC) because of its role in valuing social information. According to the results of this research, an unstable family
environment leads to low social information use later in life. There were no significant correlations found with the ACC grey matter volume or connectivity, or whole-brain connectivity. This research, therefore, shows that individual differences in social learning strategies may be formed during the environment of development, highlighting the importance of the environment for cognitive development. However, future research is necessary to further investigate the nature of the influence of the environment of development on social learning strategies and to find the neural mechanisms behind the individual differences in social learning strategies.
Keywords: social learning, social information, anterior cingulate cortex, environment of development, gACC, grey matter volume, white matter connectivity
3 The environment of development is related to individual differences in social learning strategies: an
MRI study
Individuals are making decisions all the time, deciding what to spend money on, where to go on holiday and who to vote for. These decisions are often influenced by social information. This information is gathered from, for instance, observing another individual spending money on a new car, advice from a peer for a nice holiday location or a group of friends who all vote for the same political party. However, people differ significantly in how much they rely on social information. Moreover, the extent of social information use depends on different factors, including how certain individuals are about their decision (Morgan, Rendell, Ehn, Hoppitt, & Laland, 2012) and
environmental change (Toelch et al., 2009). The extent of social information use thus depends on different factors (Morgan et al., 2012; Toelch et al., 2009; Toelch U., 2014) (e.g. task difficulty or the environmental changes) making humans flexible in deciding whether to use social information.
Despite the ability of each individual to decide whether to use social information or not, there are consistent individual differences in social learning strategies (Molleman et al., 2014) and social information use (Toelch U., 2014). Individual differences suggest that certain individuals attribute a higher value to social information than others. For instance, individuals that grow up in collectivistic societies tend to use more social information than individualistic individuals (Toelch U., 2014). Besides attributing different values to social information, people also apply different
strategies to use social information. Where some individuals tend to be ‘Payoff-based learners’, valuing the social information based on the success of the peer, others tend to be ‘Frequency-based learners’, valuing social information based on the willingness to conform to the majority (Molleman et al., 2014). The different social learning strategies co-determine the value of social information. Although individual differences in social information use have been reported, the causes and
underlying mechanisms of these individual differences in social learning are still poorly understood. It is relevant to understand the underlying causes and mechanisms of neuronal and
developmental perspectives of individual differences in social learning (Molleman et al., 2014; Toelch U., 2014). Knowing more about how individual differences in social learning strategies are formed and expressed, may result in a better understanding of cultural evolution. For instance, a group with individuals who act with different social information strategies may be able to adapt better in
situations that require different behavioural roles (Molleman et al., 2014). Variation in social learning strategies may, therefore, be important for a whole group in order to adapt to a certain situation. Thereby, if the extent of social information use is related with a certain environment of development and differences in brain structure or connectivity, it teaches something about the adaptiveness of the human brain to an early life environment. This information is crucial in order to understand human development, as social learning plays an important role in development (Herrmann, Call,
Hernández-4 Lloreda, Hare, & Tomasello, 2007). Moreover, deficits in social learning can also lead to mental disorders (Olsson & Phelps, 2007). Altogether, understanding the causes and underlying mechanisms of consistent individual differences in social learning strategies can be helpful in order to understand disorders leading from a deficit in social learning, understanding cultural evolution, and
understanding the adaptiveness of neuronal and cognitive brain development.
In order to understand the underlying causes and mechanisms of individual differences in social information strategies, this research aims to investigate the relation between individual differences in the environment of development, structural brain volume and connectivity, and social learning strategies. To answer this research question, this paper will first briefly highlight prior research concerning the environmental influences on social learning, after which the brain and social learning related previous literature will be mentioned.
Previous research suggests that the use of social learning strategies depend on the environment of development (Noble et al., 2015). It is generally assumed that developing in a rich environment (e.g. low early life stress, high socioeconomic status), has positive effects on brain development (Noble et al., 2015). When developing in a rich, stable and caring environment, it is likely that the environment is also stable (low environmental variability). Therefore, the social information is generally reliable and this results in high social information use. In contrast, it is imaginable that developing in an unstable and stressful environment (high environmental variability) leads to social information being unreliable and reduced social information use. Indeed, previous experimental research showed that social information use relies on environmental variability, where experimentally induced environmental variability predicted social information use (Toelch et al., 2009). In line with these findings, it has been shown that rats’ social learning is impaired when they experience early life stress (Levy, Melo, Galef Jr, Madden, & Fleming, 2003; Lindeyer, Meaney, & Reader, 2013; Melo et al., 2006). Based on these findings, it is hypothesized that a rich environment of development influences individual social learning strategies, resulting in more social information use.
Prior research over the years on humans and other animals has provided not only insights in the environmental influences on social cognition, but also in the main brain structures that are involved. Research in monkeys (Rudebeck, Buckley, Walton, & Rushworth, 2006), but also in humans (Behrens et al., 2009; Behrens et al., 2008; Rilling & Sanfey, 2011; Rushworth & Behrens, 2008; Tomlin et al., 2013) suggests there is an important role for the anterior cingulate cortex (ACC) in social cognition and decision making. More specifically, earlier research pointed out to a subdivision of the ACC for decision making, implicating two parallel streams (Rushworth & Behrens, 2008); one involving the ACC sulcus (sACC), and the other involving the ACC gyrus (gACC). Both streams work with a learning parameter. However, the gyral stream is more involved in processing social
5 information, where the sulcal stream is more involved in non-social information. Moreover, in
previous research (Behrens et al., 2009; Behrens et al., 2008), the activity in the gACC reflected the value given to social information, while the activity in the sACC reflected the value given to a reward. Corresponding, monkeys with lesions in the gACC no longer showed interest in social stimuli
(Rudebeck et al., 2006). In addition, human subjects whose actions relied more on peer’s advice showed greater activity in the gACC (Behrens et al., 2008). Therefore it is hypothesized that individual differences in social learning are related to structural differences in the gACC.
The gACC is also connected with multiple brain regions of the social brain network (Apps, Rushworth, & Chang, 2016). These brain areas include the temporoparietal junction (TPJ), the dorsal medial prefrontal cortex (dmPFC), amygdala, the ventral medial prefrontal cortex (vmPFC), and the anterior insula (AI). Activity in the TPJ and AI is related to individual differences in social preferences, and connectivity within the vmPFC and the striatum plays its role in integrating monetary and social value signals (van den Bos, Talwar, & McClure, 2013). The strength of the connectivity between the gACC and those social brain areas might play a role in valuing social information, as these brain areas all play its (distinct) roles in social cognition. For instance, the insula (and the dorsolateral ACC to a lesser extent) plays an important role in the detection and adaptation of behaviour to the social alignment of the group (Tomlin et al., 2013) . In sum, based on these previous findings it is
hypothesized that individual differences in social information use may also be related to the strength of the connections of the gACC and the other social brain regions.
The relation between brain structure and connectivity, social learning strategy and environment of development will be answered based on structural MRI data, diffusion-weighted data, an experimental behavioural task that measures social information use and a questionnaire about the environment of development. The grey matter volume in the gACC and its connectivity (in the cingulum) is measured. The behavioural task, called the ‘Berlin Estimate AdjuStment Task’ (BEAST) (Molleman, Kurvers, & van den Bos, 2019), measures the extent of social information use of the individual. The ‘Risky Families Questionnaire’ (Felitti et al., 2019; Taylor, Lerner, Sage, Lehman, & Seeman, 2004) is used to measure the individual environment of development and focusses on distress in the subject’s home environment during development.
First of all, it is expected in this research that a high score on the Risky Families
Questionnaire is correlated with low social information use during the BEAST, as previous research stated that high environmental variability leads to less social information use (Toelch et al., 2009), which is in line with animal models from previous studies (Levy et al., 2003; Lindeyer et al., 2013; Melo et al., 2006). Second, it is expected that the grey matter intensity of the gACC correlates with social information use and the Risky Families score. And finally, it is expected that the cingulum connectivity intensity is correlated with social information use and the Risky Families score. These
6 correlations are expected due to the previously mentioned important role of the gACC in social learning (Behrens et al., 2009; Behrens et al., 2008; Rilling & Sanfey, 2011; Rushworth & Behrens, 2008; Tomlin et al., 2013) and its connections with social brain areas (Apps et al., 2016; van den Bos et al., 2013), and environment of development being able to influence brain development (Noble et al., 2015).
Materials and Methods
Participants. In this study 67 first-year students (mean age=20, SD=2.809) of the University of
Amsterdam participated. From these students 20 were male and 47 were female. All participants were screened in order to check if they could safely enter the MRI-scan. The nationality of the participants varied, as international students of the University of Amsterdam also participated. However, each participant could understand and speak English at a minimum level to understand the questionnaire and computer task. For participating in the experiment, participants received a
monetary reward or research credits. On top of that, participants could earn a bonus (in euros), depending on how they performed on the behavioural task. All participants signed an informed consent for participating in this study.
Procedures. The procedure of this research is ethically approved by the Ethics Review Board of the
Faculty of Social and Behavioral Sciences of the University of Amsterdam. The procedure consisted of two parts, one performed online at a chosen location, and one at the Spinoza Centre for
Neuroimaging. After participants were screened for the MRI-scan and signed the informed consent, they received an email with their subject ID and a link to the online questionnaire. Participants were asked to fill in their personal subject ID at the start of the online questionnaire. As part of the online questionnaire, participants completed the BEAST (Molleman et al., 2019). Once the participant completed the online questionnaire, an appointment was made for the MRI-scan. Before entering the MRI control room, all participants signed an Informed Consent MRI General Practitioner,
consisting of information about their general practitioner and their social security number in case of finding an abnormality in the brain. After finishing the MRI-scan, the participant received the monetary reward or research credits. Based on how well participants performed during the BEAST, participants also received a bonus in cash.
The BEAST. The computer task used for this experiment to measure the amount of social information
use was the BEAST (Molleman et al., 2019). Participants performed the BEAST online (a duration of 5 minutes) together with the questionnaire and were able to earn a bonus based on how well the participant performed. During the BEAST (see figure 1), the participant got to see a number of
7 animals for 6 seconds, after which the participant needed to give a first estimate (E1) of the total
number of animals. Afterwards, the participant got to see what another participant had estimated (X) and the participant could enter a second estimate (E2) of the number of animals that were shown.
The estimation of another participant (X) was decided using X=E1*(1+Δ) if E1 was lower than the
number of animals, and X=E1*(1-Δ) if E1 was higher than the number of animals. Δ always had a value
of 0.2. Therefore, X was always given a value that was 20% higher or lower than E1, in the direction of
the correct number of animals. If the first estimate was exactly the right amount of animals, the formula of X was chosen randomly. Participants completed 5 rounds as shown in figure 1. The amount of social information use per round (s number ofround) is calculated by the next formula:
s =E2−E1
X−E1
Exclusion for s number ofround bigger than 1, and smaller than 0 was done (Molleman et al., 2019),
resulting in some participants having less than 5 rounds. Thereby, outliers (2SD higher or lower than the mean score) were excluded from the data. Finally, the social information use per participant (S) is calculated by taking the mean social information use (s) over all rounds per participant. The bonus of the participant was decided by randomly picking one round and taking the second estimate of the participant. For every 5 animals away from the correct amount of animals, 5 cents were removed from the 1 euro bonus in total.
Figure 1. The BEAST (Molleman et al., 2019). The participant gets to see a number of animals for 6 seconds (a), after which
the participant needs to make an estimate of the total number of animals displayed (b). Then, the participant is shown how much another participant estimated, after which a second estimate can be given by the participant (c).
Risky Families Questionnaire. The questionnaire used for this experiment was the ‘Risky Families
Questionnaire’ (see appendix B for the questionnaire), based on questions from two previous studies (Felitti et al., 2019; Taylor et al., 2004). The questionnaire contained 13 questions, answered using a 5-Point Likert scale. The questionnaire is used to obtain the extent of the participant’s distress that occurred in the family environment between the age of 5 and 15. It contained questions like ‘How
8 often did a parent or other adult in the household swear at you, insult you, put you down, or act in a way that made you feel threatened?’, or ‘How often would you say there was quarrelling, arguing, or shouting between your parents?’. The questionnaire is used because the family environment is the closest environment of development of an individual. Therefore, if an individual scores high on the ‘Risky Families Questionnaire’ it predicts an environment of development with stress and therefore probably a lot of variabilities. An unloving, unstable family environment could result in unreliable social information during development. The total score for the ‘Risky Families Questionnaire’ was calculated by a sum of all given answers (for the reversed questions, answers must be reversed), where a high score means that the participant grew up in a risky family. Outliers (2SD above or below the mean score) were excluded from the data.
MRI data acquisition. For the structural imaging data and the diffusion-weighted imaging data a 3
Tesla MRI scanner (Philips Achieva DS, 32 channel head coil) was used, located at the Spinoza Centre for Neuroimaging, REC-L. The scanning procedure per participant had a duration of approximately 30 minutes, of which around 17 minutes of actual scanning. The scan included a ‘Survey’ scan, two high-resolution T1-weighted anatomical scans (0.70 × 0.81 × 0.70 mm, FOV=256 × 256 × 180 mm, matrix size=368 × 318 × 257 slices, TR=11 ms, TE=5.2 ms, parallel acquisition technique=SENSE), and a diffusion-weighted imaging scan (2.00 × 2.00 × 2.00 mm, FOV=224 × 224 × 120 mm, matrix size=112 × 112 × 60 slices, TR=2550 ms, TE=77 ms, flip angle=90˚, parallel acquisition technique=SENSE).
Analysis
Behavioural data and self report. The data-analysis is done in Rstudio (Team, 2016). For analysing
the behavioural data from the BEAST and the results of the questionnaire a correlation analysis is done in order to test if a high score of the questionnaire correlates with low social information use. First, normality is tested with Shapiro-Wilk (Royston, 1982) in Rstudio. Afterwards, a Pearson’s correlation is done in Rstudio.
Grey matter ROI analyses VBM. Pre-processing and data-analysis were executed using FSL
(Jenkinson, Beckmann, Behrens, Woolrich, & Smith, 2012). For 28 subjects extracted brains from the pre-processing pipeline of MRIQC (Esteban et al., 2017) were used for performing the analysis. Due to a delay of the Spinoza Centre to carry out the pre-processing pipeline and a lack of time, the other 39 subjects were manually converted to bids format followed by brain extraction using Bet in FSL-VBM (Smith, 2002). Afterwards, the rest of the FSL-FSL-VBM protocol (Good et al., 2001) was done in order to create a symmetric grey matter template and to register all subjects to the MNI 152 standard-space using non-linear registration (Andersson, Jenkinson, & Smith, 2007). A mask to
9 analyse only the gACC was made in FSL by using the mask for the Cingulate Gyrus anterior division from the Harvard-Oxford Cortical Structural Atlas (Desikan et al., 2006) with a threshold set at 65 (figure 3.A.). Following, randomise (Winkler, Ridgway, Webster, Smith, & Nichols, 2014) was used to perform a voxel-based correlation analysis of grey matter in the gACC. A possible correlation of the volume of the gACC with the social information use and the score of the Risky Families Questionnaire is analysed using randomise. The data analysis is corrected for possible influences of age and gender to the outcome.
Also, mean grey matter intensity of the gACC is extracted using fslstats. The mean grey matter intensity, the score of the BEAST, and the score of the questionnaire is tested for normality in Rstudio using Shapiro-Wilk. A Pearson’s correlation is done with the score of the BEAST, and the Risky Families score.
Diffusion-weighted whole-brain analyses. Pre-processing and data-analysis were executed using FSL
(Jenkinson et al., 2012). Bet (Smith, 2002) is used for brain extraction, Topup (Andersson, Skare, & Ashburner, 2003) to correct for susceptibility-induced distortion and Eddy (Andersson &
Sotiropoulos, 2016) for distortion correction and subject motion. Finally, a Dtifit was executed to check if the vectors are oriented correctly compared to the anatomy of the brain and to create a fractional anisotropy (FA) map for each subject. TBSS (Smith et al., 2006) was used in FSL (Smith et al., 2004) to analyse the FA data. Analysing FA data was done by projecting all the FA data to a mean FA skeleton, after which voxelwise cross-subject statistics were applied. After TBSS, randomise (Winkler et al., 2014) was used to analyse a possible correlation of FA values in the whole brain to social information use (as result from the BEAST) and the score from the Risky Families
Questionnaire. The data analysis is corrected for possible influences of age and gender to the outcome.
Diffusion-weighted ROI analyses. After the exploratory analysis, the same correlation analysis has
been done using a region of interest mask of the cingulum. The mask of the cingulum (figure 3.B.) was created by combining the right and left cingulum from the JHU ICBM-DTI-81 White-Matter Labels atlas (Mori, Wakana, Van Zijl, & Nagae-Poetscher, 2005) provided by FSL. Afterwards, the mask was multiplied with the mean FA skeleton of all subjects to make the mask fit adequately. The mask was also used to extract the mean FA intensity using fslstats. The mean FA intensity of the cingulum, the score of the BEAST and the score of the questionnaire is tested for normality in Rstudio with Shapiro-Wilk. Afterwards, a Pearson’s correlation is performed in Rstudio with the mean FA intensity of the cingulum, the score of the BEAST and the Risky Families score.
Freesurfer grey matter analyses. From the output of the pre-processing pipeline of MRIQC (Esteban
et al., 2017) grey matter volumes of the ACC (figure 3.C.) were extracted using the aseg.stats file (Fischl et al., 2002) provided by the MRIQC pipeline. The extracted value of the ACC was analysed
10 using Rstudio. The data was tested for normality with a Shapiro-Wilk test. A Pearson’s correlation is done with the ACC grey matter volume, the data resulting from the BEAST and the data resulting from the Risky Families Questionnaire. Also, a multiple regression analysis is done to check if ACC grey matter volume, the score resulting from the Risky Families Questionnaire, gender and age predict social information use as a result of the BEAST.
Results
Behavioural data and self report. From the data of the BEAST (mean=0.209, SD=0.153) and the Risky
Families Questionnaire (mean=27.522, SD=8.663) 5 subjects were excluded due to outliers. Therefore, the analysis was performed with 62 subjects. The data resulting from the BEAST (mean=0.188, SD=0.136, W=0.938, p<0.05) and the Risky Families Questionnaire (mean=26.419, SD=7.405, W=0.938, p<0.05) were both not normally distributed (see appendix A.1. and A.2.). A Pearson’s correlation was performed, which resulted in a significant negative correlation (ρ=-0.314, t=-2.566, df=60, p=0.006, α=0.05) between social information use (S), following from the BEAST, and the Risky Families score from the questionnaire (figure 2).
Figure 2. Correlation between social information use and the environment of development. A significant negative
correlation (p<0.01, α=0.05) between social information use (S) (mean=0.188, SD=0.136), resulting from the BEAST, and the Risky Families score (mean=26.419, SD=7.405).
11
A
B
C
Figure 3. Masks used for MRI data analyses. A. The mask of the gACC in red, used for the grey matter
ROI analyses. B. The mask of the cingulum in red, used for the diffusion-weighted ROI analyses. C. The ROI of the ACC in blue, used for the grey matter analyses with the data from Freesurfer.
Grey matter ROI results. For the grey matter analyses of the gACC, 3 participants were excluded
from the data due to trouble during the pre-processing of the MRI data (corrupted data). Therefore, 59 subjects were used for grey matter ROI analyses. The correlation analyses of the gACC volume, performed using randomise, between social information use, and between the Risky Families score, were not significant (p>0.05, α=0.05). The mean grey matter intensities of the gACC (mean=0.554, SD=0.077) is normally distributed (W=0.971, p=0.175, α=0.05) (see appendix A.3.). The data resulting from the BEAST (W=0.932, p<0.01) and the Risky Families Questionnaire (W=0.939, p<0.01) were not normally distributed. A Pearson’s correlation resulted in no significant correlation (ρ=0.091, t=0.687, df=57, p=0.495, α=0.05) between the social information use (S) and the mean grey matter intensity of the gACC (figure 4.A.). Likewise, between the Risky Families score and the grey matter intensity of the gACC was no significant correlation (ρ=-0.187, t=-1.437, df=57, p=0.156, α=0.05) found (figure 4.B.).
12
A
B
Figure 4. A. Correlation between social information use and mean grey matter intensity of the gACC. No significant
correlation (p>0.05, α=0.05) was found between social information use (S) and mean grey matter intensity of the gACC (mean=0.554, SD=0.077). B. Correlation between the environment of development and mean grey matter intensity of
the gACC. No significant correlation (p>0.1, α=0.05) was found between the Risky Families score and mean grey matter
intensity of the gACC (mean=0.554, SD=0.077).
Diffusion-weighted whole-brain and ROI results. For analysing the diffusion-weighted data, 62
subjects were used. The correlation analyses of the whole brain and the cingulum, performed using randomise, between social information use, and between the Risky Families score, were not
significant (p>0.05, α=0.05). The data of the mean fractional anisotropy (FA) intensity of the cingulum (mean=0.557, SD=0.027) was normally distributed (W=0.981, p=0.437, α=0.05) (see appendix A.4.). The data of the BEAST (W=0.938, p<0.01) and the Risky Families Questionnaire (W=0.938, p<0.01) were not normally distributed. There was no significant correlation (ρ=0.153, t=1.202, df=60, p=0.234, α=0.05) found between the mean FA intensity of the cingulum and the social information use (S) (figure 5.A.). Likewise, no significant correlation (ρ=-0.038, t=-0.291, df=60, p=0.772, α=0.05) was found between the mean FA intensity of the cingulum and the Risky Families score (figure 5.B.).
13
A
B
Figure 5. A. Correlation between social information use and mean functional anisotropy (FA) intensity of the cingulum.
No significant correlation (p>0.1, α=0.05) was found between social information use (S) and mean FA intensity of the cingulum (mean=0.557, SD=0.027). B. Correlation between the environment of development and mean functional
anisotropy (FA) intensity of the cingulum. No significant correlation (p>0.05, α=0.05) was found between the Risky
Families score and mean FA intensity of the cingulum (mean=0.557, SD=0.027).
Freesurfer grey matter ROI results. For analysing the data of the ACC grey matter volume resulting
from Freesurfer stats, 61 subjects were used. The data of the ACC volume (mean=857.202 voxels, SD=158.365 voxels) was not normally distributed (W=0.960, p<0.05) (see appendix A.5.). Also, the data of the BEAST (W=0.935, p<0.01) and the Risky Families Questionnaire (W=0.933, p<0.01) were not normally distributed. No significant correlation (ρ=-0.059, t=-0.453, df=59, p=0.652, α=0.05) was found between ACC volume and the social information use (S) (figure 6.A.), and no significant
correlation (ρ=0.058, t=0.447, df=59, p=0.657, α=0.05) was found between ACC volume and the Risky Families score (figure 6.B.). The model (social information use (S) predicted by ACC grey matter volume, Risky Families score, gender, and age) tested with a multiple regression analysis (table 1) was not significant (F=2.026, df1=4, df2=56, p=0.103, α=0.05). Of all predictors, only the Risky Families score was a significant predictor significant (p=0.018, α=0.05).
14
A
B
Figure 6. A. Correlation between social information use and grey matter volume of the ACC. No significant correlation
(p>0.1, α=0.05) was found between social information use (S) and the grey matter volume of the ACC (mean=857.202 voxels, SD=158.365 voxels). B. Correlation between the environment of development and grey matter volume of the
ACC. No significant correlation (p>0.1, α=0.05) was found between the Risky Families score and grey matter volume of
the ACC (mean=857.202 voxels, SD=158.365 voxels).
B t-value p-value
ACC gm volume -2.538*10-5 -0.234 0.81608
Risky Families score -5.675 *10-3 -2.433 0.01820 *
Gender -2.308*10-3 -0.124 0.90200
Age -7.360*10-3 -1.236 0.22163
Table 1. Multiple regression analysis for predicting social information use. The output table of the multiple regression
analysis to predict social information use (S). Predictors are ACC grey matter, the Risky Families score, gender and age. Only Risky Families score* was a significant predictor for social information use (S) (p<0.05, α=0.05). Also the estimate of the regression is given (B) and the t-value.
Discussion
The main goal of this research was to investigate the relation between individual differences in social learning strategies, the environment of development, and anterior cingulate cortex volume and connectivity. The social learning strategy was measured by the extent of social information use, resulting from the BEAST. The environment of development was measured using a score, resulting from the Risky Families Questionnaire. A significant negative correlation is found between social information use and the score resulting from the Risky Families Questionnaire. No significant correlation was found between ACC grey matter volume, gACC grey matter intensity, or gACC FA intensity and social information use, or the environment of development. Also, whole-brain connectivity analysis showed no significant correlation with either social information use or the environment of development. These results support the hypothesis that a rich environment of development influences social learning, resulting in more social information use. However, the
15 results do not support the hypothesis that social learning is related with structural differences in gACC volume and connectivity.
It is likely to conclude, based on these results, that developing in an unstable environment relates to a cognitive adaptation that decreases the use of social information throughout life. These results are in line with prior animal research that showed impaired social learning when developing in a stressful environment (Levy et al., 2003; Lindeyer et al., 2013; Melo et al., 2006), and
environmental variability reducing social information use in humans (Toelch et al., 2009). If social information provided by the environment is constantly unreliable to learn from, it is for the better to adopt a learning strategy that does not rely on social information. This way, the early life experience of the value of social information could shape the value of social information later in life. As social learning plays an important role in cognitive development (Herrmann et al., 2007), the found long-term influence of the environment of development on social learning strategies is of great interest. Moreover, these findings contribute to understanding cultural evolution, as a group with different social learning strategies may be better to adapt in multiple varieties of situations (Molleman et al., 2014). Also, the development of social learning strategies can have an impact on mental disorders leading from a deficit in social learning (Olsson & Phelps, 2007). The results of this research may therefore contribute to understanding the necessity of the environment of development for cognitive development.
Based on the results, it seems that the social learning strategy is a fixed effect of the environment of development, which could cause implications. It is beneficial to rely on a social learning strategy that is learned by the early environment of development if the reliability of social information stays the same during life. However, the reliability of social learning could switch later in life. If the reliability of social information switches later in life, it is unbeneficial to stay with the same social learning strategy. Sticking with the same social learning strategy, while the environment has changed, may result in individuals mismatching a value to social information. This way, individuals can not use the optimum learning strategy in the given situation. Moreover, as social learning can influence everyday decisions as who to vote for or what to spend money on, a nonbeneficial learning strategy can influence important decisions made in life negatively. To investigate if social learning strategies are a fixed effect of the environment of development, future research in individual differences in social learning strategies could take into account the current environment of development and how the environment has changed over life.
Unexpected, the results that no significant neural correlations were found with the MRI data are not in line with the expectation that it is likely to find a neural correlation with social information use, or the environment of development, which is based on prior research. Prior research stated that the environment of development also influences brain development (Noble et al., 2015) and has
16 linked gACC activity with valuing social information (Behrens et al., 2009; Behrens et al., 2008; Rilling & Sanfey, 2011; Rushworth & Behrens, 2008; Tomlin et al., 2013), implying an important role for the gACC in the individual differences in social learning strategies. Based on the important role given to the gACC in social learning in prior research, it is unexpected to find no correlation at all between social information use and structural differences in gACC connectivity and volume. The unexpected results could have several possible explanations.
One explanation for not finding any structural brain differences in this study could be that the behavioural task is limited to only measure social information use, which does not exactly distinguish all social learning strategies (Mesoudi, Chang, Dall, & Thornton, 2016). Two different social learning strategies can both result in social information use. For example, some individuals rely on social learning when they are uncertain about their decision, other individuals rely on social learning to conform to social alignment. Moreover, if participants entirely copy social information half of the time and not copy at all the other half of the time, the score of the BEAST would result in the same amount of social information use for someone always copying 50% of the social
information. It is thus possible that two individuals have the same social information use score (S) resulting from the BEAST, while both acting based on a different motivation. That different motivation may manifest in different neural correlates inducing the value of social information. Therefore, two individuals who both use social information could have different structural correlates in the brain inducing the social information use while matching a different social learning strategy. If this may be the case, it could explain why no correlates between social information use and neural structures and connectivity are found in this research.
Another explanation for not finding any brain-related correlations with social information use could be that the neural foundation of individual differences in social learning does not simply derive from a structural difference in one brain area. As mentioned before, the gACC is connected with many other social brain areas (Apps et al., 2016), contributing to the process of social learning. The individual differences in social learning could be, for instance, an effect of a specific combination of connectivity and grey matter volume in multiple brain structures at the same time. This way, the gACC activity linked to social information use found in prior research (Behrens et al., 2009; Behrens et al., 2008; Rilling & Sanfey, 2011; Rushworth & Behrens, 2008; Tomlin et al., 2013) could be induced by a sum of brain areas and its connections with the gACC, suggesting a more complex process underlying the individual differences in social learning, that does not express itself in structural differences in grey matter or connectivity in the gACC. This might be a reason why no correlates are found in gACC grey matter and connectivity in this research.
Future research is necessary to find the, still unknown, neural correlates of the individual differences in social learning. As the consistent differences in social information use may be
17 predicted by the environment of development, and the environment of development also influences brain development (Noble et al., 2015), it is still likely that there are neural correlates of the
individual differences in social learning. However, a different approach is necessary to find those neural correlates, taking into account a more complex process underlying social information use. Instead of analysing just gACC volume and connectivity, future research must take into account the connectivity strength of the gACC with other social brain areas. This way, more focus is on analysing the total neural network involving social information use. On top of that, future research must also check for different social learning strategies within social information use. This may be done after completing the BEAST by a self-report about the motive to use social information. By checking the motive behind social information use, possible different neural correlates of different social learning strategies that all result in social information use can be taken into account.
Altogether, the results of this research suggest that the environment of development contributes to individual differences in social learning strategies throughout life. According to the results of this research, developing in an unstable environment may reduce social information use later in life. These findings are broadening the knowledge about human cognitive development and highlight the importance of the environment of development because of its long-lasting effects. This adaptive ability of humans to the environment of development may not only play its role in individual development but may also play its role in cultural evolution. Thereby, this research contributes to the understanding of the underlying causes of individual differences in social learning strategies.
However, the neural causes and underlying mechanisms are still unknown and require more
research. As social learning plays an important role in cognitive development (Herrmann et al., 2007), this future research is relevant in order to deeper understand social learning and disorders leading from a deficit in social learning.
18
Literature
Andersson, J. L., Jenkinson, M., & Smith, S. (2007). Non-linear registration aka Spatial normalisation FMRIB Technial Report TR07JA2. FMRIB Analysis Group of the University of Oxford.
Andersson, J. L., Skare, S., & Ashburner, J. (2003). How to correct susceptibility distortions in spin-echo spin-echo-planar images: application to diffusion tensor imaging. Neuroimage, 20(2), 870-888.
Andersson, J. L., & Sotiropoulos, S. N. (2016). An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging. Neuroimage, 125, 1063-1078. Apps, M. A., Rushworth, M. F., & Chang, S. W. (2016). The anterior cingulate gyrus and social
cognition: tracking the motivation of others. Neuron, 90(4), 692-707.
Behrens, T. E., Hunt, L. T., & Rushworth, M. F. (2009). The computation of social behavior. Science, 324(5931), 1160-1164.
Behrens, T. E., Hunt, L. T., Woolrich, M. W., & Rushworth, M. F. (2008). Associative learning of social value. Nature, 456(7219), 245.
Desikan, R. S., Ségonne, F., Fischl, B., Quinn, B. T., Dickerson, B. C., Blacker, D., . . . Hyman, B. T. (2006). An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage, 31(3), 968-980.
Esteban, O., Birman, D., Schaer, M., Koyejo, O. O., Poldrack, R. A., & Gorgolewski, K. J. (2017). MRIQC: Advancing the automatic prediction of image quality in MRI from unseen sites. PloS one, 12(9), e0184661.
Felitti, V. J., Anda, R. F., Nordenberg, D., Williamson, D. F., Spitz, A. M., Edwards, V., . . . Marks, J. S. (2019). Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: The Adverse Childhood Experiences (ACE) Study. American journal of preventive medicine, 56(6), 774-786.
Fischl, B., Salat, D. H., Busa, E., Albert, M., Dieterich, M., Haselgrove, C., . . . Klaveness, S. (2002). Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron, 33(3), 341-355.
Good, C. D., Johnsrude, I. S., Ashburner, J., Henson, R. N., Friston, K. J., & Frackowiak, R. S. (2001). A voxel-based morphometric study of ageing in 465 normal adult human brains. Neuroimage, 14(1), 21-36.
Herrmann, E., Call, J., Hernández-Lloreda, M. V., Hare, B., & Tomasello, M. (2007). Humans have evolved specialized skills of social cognition: The cultural intelligence hypothesis. Science, 317(5843), 1360-1366.
Jenkinson, M., Beckmann, C. F., Behrens, T. E., Woolrich, M. W., & Smith, S. M. (2012). Fsl. Neuroimage, 62(2), 782-790.
Levy, F., Melo, A., Galef Jr, B., Madden, M., & Fleming, A. (2003). Complete maternal deprivation affects social, but not spatial, learning in adult rats. Developmental Psychobiology: The Journal of the International Society for Developmental Psychobiology, 43(3), 177-191. Lindeyer, C. M., Meaney, M. J., & Reader, S. M. (2013). Early Maternal Care Predicts Reliance on
Social Learning About Food in Adult Rats. Developmental Psychobiology, 55.2, 168-175. doi:10.1002/dev.21009
Melo, A. I., Lovic, V., Gonzalez, A., Madden, M., Sinopoli, K., & Fleming, A. S. (2006). Maternal and littermate deprivation disrupts maternal behavior and social‐learning of food preference in adulthood: Tactile stimulation, nest odor, and social rearing prevent these effects.
Developmental Psychobiology: The Journal of the International Society for Developmental Psychobiology, 48(3), 209-219.
Mesoudi, A., Chang, L., Dall, S. R., & Thornton, A. (2016). The evolution of individual and cultural variation in social learning. Trends in ecology evolution, 31(3), 215-225.
Molleman, L., Kurvers, R., & van den Bos, W. (2019). Unleashing the BEAST: a brief measure of human social information use. Evolution and Human Behavior.
19 Molleman, L., Van den Berg, P., & Weissing, F. (2014). Consistent individual differences in human
social learning strategies. Nature Communications, 5, 3570.
Morgan, T. J., Rendell, L. E., Ehn, M., Hoppitt, W., & Laland, K. N. (2012). The evolutionary basis of human social learning. Proceedings of the Royal Society B: Biological Sciences, 279(1729), 653-662. doi:10.1098/rspb.2011.1172
Mori, S., Wakana, S., Van Zijl, P. C., & Nagae-Poetscher, L. (2005). MRI atlas of human white matter: Elsevier.
Noble, K. G., Houston, S. M., Brito, N. H., Bartsch, H., Kan, E., Kuperman, J. M., . . . Libiger, O. (2015). Family income, parental education and brain structure in children and adolescents. Nature neuroscience, 18(5), 773.
Olsson, A., & Phelps, E. A. (2007). Social learning of fear. Nature neuroscience, 10(9), 1095. Rilling, J. K., & Sanfey, A. G. (2011). The neuroscience of social decision-making. Annual review of
psychology, 62, 23-48.
Royston, J. (1982). Algorithm AS 181: the W test for normality. Journal of the Royal Statistical Society. Series C, 31(2), 176-180.
Rudebeck, P. H., Buckley, M. J., Walton, M. E., & Rushworth, M. F. (2006). A role for the macaque anterior cingulate gyrus in social valuation. Science, 313(5791), 1310-1312.
Rushworth, M. F., & Behrens, T. E. (2008). Choice, uncertainty and value in prefrontal and cingulate cortex. Nature neuroscience, 11(4), 389.
Smith, S. M. (2002). Fast robust automated brain extraction. Human brain mapping, 17(3), 143-155. Smith, S. M., Jenkinson, M., Johansen-Berg, H., Rueckert, D., Nichols, T. E., Mackay, C. E., . . .
Matthews, P. M. (2006). Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. Neuroimage, 31(4), 1487-1505.
Smith, S. M., Jenkinson, M., Woolrich, M. W., Beckmann, C. F., Behrens, T. E., Johansen-Berg, H., . . . Flitney, D. E. (2004). Advances in functional and structural MR image analysis and
implementation as FSL. Neuroimage, 23, S208-S219.
Taylor, S. E., Lerner, J. S., Sage, R. M., Lehman, B. J., & Seeman, T. E. (2004). Early environment, emotions, responses to stress, and health. Journal of personality, 72(6), 1365-1394. Team, R. (2016). RStudio: Integrated Development Environment for R. RStudio, Inc., Boston, MA. Toelch, U., van Delft, M. J., Bruce, M. J., Donders, R., Meeus, M. T., & Reader, S. M. (2009). Decreased
environmental variability induces a bias for social information use in humans. Evolution and Human Behavior, 30(1), 32-40.
Toelch U., B. M. J., Newson L., Richerson P.J., Reader S.M. (2014). Individual consistency and flexibility in human social information use. Proceedings of the Royal Society, 281.
doi:http://dx.doi.org/10.1098/rspb.2013.2864
Tomlin, D., Nedic, A., Prentice, D. A., Holmes, P., & Cohen, J. D. (2013). The neural substrates of social influence on decision making. PloS one, 8(1), e52630.
van den Bos, W., Talwar, A., & McClure, S. M. (2013). Neural correlates of reinforcement learning and social preferences in competitive bidding. Journal of Neuroscience, 33(5), 2137-2146.
Winkler, A. M., Ridgway, G. R., Webster, M. A., Smith, S. M., & Nichols, T. E. (2014). Permutation inference for the general linear model. Neuroimage, 92, 381-397.
20
Appendices Appendix A. Distributions of the variables.
A.1. Distribution of the BEAST. The distribution of social
information use (S) (mean=0.188, SD=0.136) was not normally distributed (W=0.938, p<0.05).
A.2. Distribution of the Risky Families Questionnaire. The
distribution of the Risky Families score (mean=26.419, SD=7.405) was not normally distributed (W=0.938, p<0.05).
A.3. Distribution of the grey matter intensity of the gACC.
The distribution of the mean grey matter intensity (mean=0.554, SD=0.077) was normally distributed (W=0.971, p=0.175, α=0.05).
A.4. Distribution of the fractional anisotropy (FA) intensity of the cingulum. The distribution of the mean FA
intensity of the cingulum (mean=0.557, SD=0.027) was normally distributed (W=0.981, p=0.437, α=0.05).
21
A.5. Distribution of the grey matter volume (voxels) of the ACC. The distribution of the grey matter volume of the ACC
(mean=857.202 voxels, SD=158.365 voxels) was not normally distributed (W=0.960, p<0.05).
Appendix B. Questions of Risky Families Questionnaire.
These are questions about when you were growing up (between ages 5 – 15). Please think over your family life and answer these questions.
1. How often did a parent or other adult in the household make you feel that you were loved, supported, and cared for?
Not at All 1 2 3 4 5 Very Often
2. How often did a parent or other adult in the household swear at you, insult you, put you down, or act in a way that made you feel threatened?
Not at All 1 2 3 4 5 Very Often
3. How often did a parent or other adult in the household express physical affection for you, such as hugging, or other physical gestures of warmth and affection?
Not at All 1 2 3 4 5 Very Often
4. How often did a parent or other adult in the household push, grab, shove, or slap you?
22
5. How often would you say that a parent or other adult in the household behaved violently toward a family member or visitor in your home?
Not at All 1 2 3 4 5 Very Often
6. How often would you say there was quarrelling, arguing, or shouting between your parents?
Not at All 1 2 3 4 5 Very Often
7. How often would you say there was quarrelling, arguing, or shouting between a parent and you?
Not at All 1 2 3 4 5 Very Often
8. How often would you say there was quarrelling, arguing, or shouting between a parent and one of your siblings?
Not at All 1 2 3 4 5 Very Often
9. How often would you say there was quarrelling, arguing, or shouting between your sibling(s) and you?
Not at All 1 2 3 4 5 Very Often
10. Would you say the household you grew up in was chaotic and disorganized?
Not at All 1 2 3 4 5 Very Often
11. In your childhood, did you live with anyone who was a problem drinker or alcoholic, or who used street drugs?
23
Not at All 1 2 3 4 5 Very Often
12. Would you say that the household you grew up in was well-organized and well managed?
Not at All 1 2 3 4 5 Very Often
13. Would you say you were neglected while you were growing up, left on your own to fend for yourself?
