“Wanting” versus “needing” related value: An fMRI meta‐analysis

Abstract Consumption and its excesses are sometimes explained by imbalance of need or lack of control over “wanting.” “Wanting” assigns value to cues that predict rewards, whereas “needing” assigns value to biologically significant stimuli that one is deprived of. Here we aimed at studying how the brain activation patterns related to value of “wanted” stimuli differs from that of “needed” stimuli using activation likelihood estimation neuroimaging meta‐analysis approaches. We used the perception of a cue predicting a reward for “wanting” related value and the perception of food stimuli in a hungry state as a model for “needing” related value. We carried out separate, contrasts, and conjunction meta‐analyses to identify differences and similarities between “wanting” and “needing” values. Our overall results for “wanting” related value show consistent activation of the ventral tegmental area, striatum, and pallidum, regions that both activate behavior and direct choice, while for “needing” related value, we found an overall consistent activation of the middle insula and to some extent the caudal‐ventral putamen, regions that only direct choice. Our study suggests that wanting has more control on consumption and behavioral activation.

interoceptive salience (Craig, 2003). "Wanting" is related to the prediction of reward in the brain and is more closely related to motivation (Berridge, 2004). Although earlier theories suggested that need or deprivation defines motivation (Hull, 1943), it was later demonstrated that cues that signal hunger do not elicit motivation to eat, while foodrelated cues did lead to motivation to eat (Bindra, 1974), showing that motivated behaviors are more determined by reward prediction (which is closer to "wanting") than need state, though the latter can have a multiplicative effect (Toates, 1994). The distinction might seem clear. However, since "needed" stimuli are often pursued and thus associated with motivational value, it is not obvious whether an external stimulus, such as food, is pursued because of its "needing" or "wanting" related value. As "needed" stimuli might have a different form of value than "wanted" stimuli, we can wonder how in the brain the value of "needed" stimuli differs from that of "wanted" stimuli.
Given that we cannot test wanting and/or needing as general phenomena, here we are testing some manifestation of them. Hence, to conceptualize this distinction, we refer to "wanting a stimulus" as "Wanting" ST , which represents a brain reaction to a reward predicting cue that triggers reward seeking. We refer to "needing a stimulus" as "Needing" ST , which represents the brain reaction to a stimulus that one is deprived of, without necessarily seeking it. We propose to use activation pattern to the perception of a cue predicting a reward as a model for "Wanting" ST related value, and to use the activation pattern during the perception of a food stimulus while in the state of hunger as a model for "Needing" ST related value (Silverman et al., 2014;Spear, 2011). In order to answer our question, we used a neuroimaging meta-analytic approach, comparing the patterns of brain activations during the processing of "Wanting" ST versus "Needing" ST . Previous meta-analyses have focused on either "Wanting" ST (Oldham et al., 2018;Sescousse et al., 2013;Wilson et al., 2018) or "Needing" ST (Chen & Zeffiro, 2020;LaBar et al., 2001;van der Laan et al., 2011), but no work has directly compared both activation patterns.

"Wanting" ST related value
Nonhuman animals tend to respond and "want" food even when they are no longer hungry (Bouton, 2016). This is also the case for humans: cues of M&M or pictures of cigarettes (for smokers) lead to more consumption even after having been consumed to satiety (Hogarth & Chase, 2011;Watson et al., 2014). "Wanting" ST is a concept from incentive salience theory that comes from animal studies (Berridge & Robinson, 1998; see also : Ikemoto & Panksepp, 1999;Salamone & Correa, 2002;Salamone et al., 1997), which states that "Wanting" ST is based on two neuropsychological processes: the first is a pavlovian cue that predicts the reward, and the second is the dopaminergic state (which might be enhanced by hunger, thirst, emotions, drugs, etc.) (Berridge, 1996). In other words, "Wanting" ST depends on external stimuli that act as pavlovian cues that predict rewards (Berridge, 2018). The attribution of value to these cues depends on mesolimbic dopamine (Berridge, 1996). The latter is secreted within the ventral tegmental area (VTA) by a reward cue (Schultz, 1998) and projected within the nucleus accumbens (Nacc), ventral pallidum, and on the central amygdala opioid (Warlow & Berridge, 2021;Zhang et al., 2009).
Hence, hunger can cautiously be used as a proxy for "Needing" ST . Thus, based on the inherent association between hunger (and thirst) and "Needing" ST , the insula and ACC, and to some extent the OFC, amygdala, and caudate may be engaged in the processing of "Needing" ST related value and contribute to directional value.

Two types of predictions and values
The conceptualization of "Wanting" ST and "Needing" ST here as processing of either wanted or needed stimuli implies two forms of predictions. Indeed, in both cases, the value of stimuli often depends on a prediction and a prediction error: in case of "Wanting" ST , that prediction is related to reward (unexpected reward or reward predicting cue) and is computed in the ventral striatum (Takahashi et al., 2016), while the prediction error is computed in the VTA (Schultz, 2015).
For "Needing" ST , that prediction is related to interoception (predicted state vs. sensed state) and is said to be computed in visceromotor cortices (OFC, ACC, anterior insula), whereas the interoceptive prediction error is proposed to be computed within the mid-posterior insula (Barrett & Simmons, 2015). Moreover, both "Wanting" ST and "Needing" ST establish a relation between the state ("wanting" state and "needing" state) and some external stimuli, where the state attributes some form of value to the stimuli. The value assigned to stimuli (by both "Wanting" ST and "Needing" ST ) can have a directional effect or activational effect. The directional effect is linked to choice (preference or action selection) and directs towards or away from stimuli, while the activational effect is related to action and its initiation, maintenance, and vigor or effort (see Salamone et al., 2018). Indeed, "Wanting" ST depends on mesolimbic dopamine (Berridge, 1996) which provides full motivational value to reward as it provides both activational value (or effect) and directional value (or effect) to stimuli (see Salamone et al., 2018). "Needing" ST (by itself) does not seem to provide the activational value that "Wanting" ST provides to stimuli (Berridge, 2004; see also Salamone et al., 2018). However, "Needing" ST does provide directional value (Balleine, 1992;Salamone et al., 2018).
Importantly, if both "Wanting" ST and "Needing" ST provide directionalrelated value to stimuli that impacts choice, they do so in different ways. In the case of "Wanting" ST , that choice value is pavlovian cue triggered and stimulus related (Balleine, 2009;Berridge, 2012). In the case of "Needing" ST , that choice value is act-outcome based (Balleine, 2009). Hence, depending on either state, stimulus value could be represented by different activation patterns in the brain Dayan & Balleine, 2002). However, no work has quantitatively tested this hypothesis.
Although "Wanting" ST takes both cues and physiological states, and the latter can be related to need states (Berridge, 1996;Zhang et al., 2009), not all physiological states are related to needs, some are related to emotions, drugs, and so forth (Berridge, 1996;Zhang et al., 2009). Moreover, need states can give and control value to relevant rewards in relation to their outcome by associating the discriminative properties of needs with the increased value placed on the reward (Balleine, 1992;Nader et al., 1997), and such "directing" motivational control through determination of specific outcome values is said to be dopamine independent (Balleine, 2005;Niv et al., 2006; see also Salamone et al., 2018). We believe there is an objective reaction to stimuli that are needed, even in absence of an incentive salience motivation, that is, "Wanting" ST . In the same way, there is an objective reaction to "Liking" ST in absence of "Wanting" ST (see Berridge, 1996); and in the same way, there is an objective "Wanting" ST , that can happen regardless of "Needing" ST or "Liking" ST (see Berridge, 2004). Regarding the difference between "Wanting" ST versus "Needing" ST , they likely provide two different "roles" (or values) to reward cues. We believe that, depending on situations or paradigms, "Wanting" ST related cues are stimuli that motivate action and bias behavior (see Robinson et al., 2014); whereas the "Needing" ST related cues are stimuli that are outcome relevant for the current need state (see Balleine, 2005), and our view is that those two roles/values do not necessarily apply at the same time to stimuli depending on situations/paradigms. Our study is thus about when needing a stimulus influences its processing without the wanting component for that stimulus, and when "Wanting" ST happens without "Needing" ST .
As discussed, neuroimaging studies in the activation pattern of "Wanting" ST related value shows consistent activation of the striatum, amygdala, and insula (Carter et al., 2009;Knutson et al., 2001Knutson et al., , 2003O'Doherty, 2004;O'Doherty et al., 2002;Oldham et al., 2018;Schott et al., 2008;Wilson et al., 2018). Those same regions have also been found in the activation pattern of "Needing" ST (Führer et al., 2008;Goldstone et al., 2009;LaBar et al., 2001;Mohanty et al., 2008;Siep et al., 2009;van der Laan et al., 2011). It is not clear how these regions contribute to either "Wanting" ST or "Needing" ST . Our goal is thus to use a meta-analytic approach to compare the consistent brain activation patterns for "Wanting" ST and "Needing" ST related values by identifying similarities and differences between the brain activation patterns of these two states that guide value attribution and our consumption behaviors. To do this, we will quantitatively identify the consistent activation patterns after the observation of a reward cue/reward prediction ("Wanting" ST ), versus while (or after) observing a food cue when hungry ("Needing" ST ). We will then directly compare these activation patterns by conducting meta-analytic conjunction and contrast analyses.

Yes Yes
Note: The red colored "yes" means the criterion is crucial for the definition of either "Wanting" ST or "Needing" ST . Most mean BMI values were below 30, which is the threshold of obesity as defined by the World Health Organization (World Health Organization, 2020); however, we took people that had up to 30-35 of BMI, especially in contrast when healthy and overweight were mixed. For the Millman et al.'s (2020) study, we took the contrast of "large gain > small gain" (so exceptionally, we contrasted two rewards); first because there were no contrast for gain alone in general, and because small gains as well as large losses were received when participants failed to respond within the allowable time window, so these two outcomes served as de facto negative RPEs. Abbreviation: MRI, magnetic resonance imaging.

METHODOLOGY
We decided to use a meta-analytic approach as it provides an opportunity to quantitatively assess brain activation patterns of "Wanting" ST versus "Needing" ST related values using large collections of data. This is useful as a summary of the existing literature is needed, not just because both concepts have rarely been directly compared and are often studied separately in neuroimaging studies, but also because each study might have low replicability, analytical and experimental flexibility, and/or small samples. Thus, our approach aims at identifying and comparing regions that are consistently activated for "Wanting" ST and those that are consistently activated for "Needing" ST .
Specifically, we first conducted two meta-analyses to quantitatively summarize results from fMRI published studies on the reward prediction for "Wanting" ST (activation maps taken when participants received a reward predicting cue that triggers reward seeking); and on perceiving food stimulus while being hungry for "Needing" ST (activation maps taken when participants perceived food while hungry).
Second, we did a conjunction analysis to identify common regions that are consistently activated in both states. Finally, we contrasted "Wanting" ST and "Needing" ST consistent activation patterns by testing, ["Wanting" ST -"Needing" ST ], and ["Needing" ST -"Wanting" ST ].

Included articles
Based on the view that "Wanting" ST rests upon reward prediction that has been turned into a decision (see ), we used the following keywords to identify articles related to "Wanting" ST : (("prediction" AND "anticipation") OR "desire" OR "wanting") Based on the view that "Needing" ST such as hunger depends on interoception (Craig, 2003) coming from deprivation of something biologically important, we used the following keywords to identify articles related to "Needing" ST : ("alliesthesia" OR "interoceptive" OR "loss aversion" OR "need" OR "homeostasis" OR "modulating factor" OR "self-specificity" OR "self-referential" OR "hunger" OR "food deprivation").
While the previous lists of keywords were specific to either "Wanting" ST or "Needing" ST , the following keywords were the same for both "Wanting" ST and "Needing" ST ; those keywords were the following: ("reward" OR "motivation" OR "goal directed" OR "decision-making" OR "seeking" OR "incentive") AND ("fMRI") These include words that are often conceptualized as related to "Wanting" ST and to "Needing" ST (Bouton, 2016;Panksepp, 2004).
For both "Wanting" ST or "Needing" ST , the following inclusion criteria were used: healthy subjects; whole-brain analyses (with or without    [ALE] using Lancaster transform); maps were corrected (or cluster level corrected); activation contrast only.
With regard to "Wanting" ST , we typed the keywords on PubMed (February 2021). The database returned 159 articles. The main selection criteria were the presence of a cue that predicts a reward and triggers reward seeking contrasted with no prediction of reward (reward prediction > no reward prediction). After evaluation based on these criteria, 19 final articles were selected out of 26 that were fully read, and from which we found three additional articles from reviews and other articles that met all the criteria for "Wanting" ST for a total of 22 selected articles (see Table 2 for list of retained articles).
Note that these rewards were mostly money or points, so they are not (directly) related to food, but they are used because "Wanting" ST or incentive motivation activates a general system, regardless of the type of stimulus (Bindra, 1968;Bouton, 2016). See PRISMA in Supporting Information for step-by-step exclusion of articles.
Regarding "Needing" ST related articles, we typed the keywords on PubMed (February 2021). The database returned 376 articles.
The main logic was to select experiments when subjects were in a hungry state and perceiving a food stimulus. We looked for both "hunger > baseline" as well as "hunger > satiety" contrasts, because of the inherent subtraction logic of fMRI and in order to have a larger number of experiments. Hence, the two main criteria were: (1) presence of a privation contrast: hunger + stimulus > satiety + stimulus, or hunger + stimulus > baseline; (2) the participant was perceiving some food stimulus which could be presented in any modality: visual, taste, odor, and so forth. Using the selection criteria (see Table 1 for all criteria), we kept 26 articles. After fully reading the final 26 articles, nine were selected, and we found some additional ones through other articles and reviews, and seven among them matched all criteria for "Needing" ST (hunger) for a total of 16 articles (see Table 3 for list of retained articles). See PRISMA and Supporting Information for step-by-step exclusion of articles.
In order to more easily disentangle "Wanting" ST versus "Needing" ST , the experiments included in "Wanting" ST did not include need states, and the "Needing" ST related experiments did not include situations in which a cue triggers behavior/reward seeking. This might be viewed as nonfasting versus fasting, but such interpretation should be taken cautiously as it is not about the same reward type, and "Wanting" ST and "Needing" ST are dependent on different situations/paradigms and reward cue roles. Of note, "Wanting" ST studies did not explicitly exclude food-related studies and were not limited to secondary rewards. However, our selection criteria resulted in the fact that we did not find food related "Wanting" ST studies to include. Hence, unintentionally, "Needing" ST included primary reward whereas "Wanting" ST included secondary rewards and motor action. For "Needing" ST , the included studies used a contrast on perception of food-related stimuli while hungry versus while satiated. Thus, brain activity elicited by food itself would cancel out, and only the "Needing" ST part should remain.
Regarding the experimental task, we share the view that "Wanting" ST is related to the preparatory and motivational excitement for motor behavior, specifically within the NAcc (Cardinal et al., 2002). Aside from that, the "Wanting" ST contrast is between anticipation of a reward cue versus anticipation of nonreward cues, where participants responded whether they were rewarded or not. In that sense, motor preparation per se would likely cancel out or would not account for all the brain regions of such contrast.

Meta-analyses
Meta-analyses were conducted with the activation likelihood estimation (ALE) approach using the Brainmap's GingerALE application.
Independently introduced by Turkeltaub et al. in order to test the differentiation between true convergence of foci and random clustering (Eickhoff et al., 2012). As a method of inference, the new algorithm uses random-effects analysis that calculates the above-chance clustering between experiments. Furthermore, the new algorithm gives more weight to gray matter compared to white matter by limiting the meta-analysis to an anatomically constrained space specified by a gray matter mask. Contrast analyses are based on two different datasets (i.e., two previous ALE results) and thus compare two different sets of foci for statistically significant differences, and the conjunction is the intersection of the thresholded maps.
In our analyses, we used the MNI152 coordinate system and the less conservative (larger) mask size. For "Wanting" ST , there were 21 articles, 34 experiments, 3306 subjects, and 572 foci (see Tables 2 and 3 for all included articles

"Wanting" ST
Our first meta-analysis was on "Wanting" ST (Table 4 and Figure 1).
This meta-analysis revealed consistent activations within the following regions: the left putamen, the left globus pallidus (which encompassed the nucleus accumbens), the left caudate body and right caudate head, the left substantia nigra, the right red nucleus (encompassing the ventral tegmental area), the right hypothalamus, the bilateral thalamus, the left precentral gyrus, the left inferior parietal lobule, the right dorsolateral and medial prefrontal cortex, the right superior parietal lobule, the right claustrum (whose cluster was mainly the anterior insula).
Of note, because we had much more MID (monetary incentive delay) tasks (see Knutson et al., 2000) in the "Wanting" ST contrast, we conducted a single meta-analysis with only studies that did not use the MID task (please see non-MID task meta-analysis in the Supporting Information), and we found peak activity within the ventral and dorsal striatum, the dopaminergic midbrain (VTA/SN), and anterior insula.
Thus, though experimental tasks for "Wanting" ST included a lot of MID tasks, other included paradigms elicited (separately from the MID tasks) the same mesolimbic dopamine and ventral striatal network that has been related to incentive salience "Wanting" ST ; although in terms of overall whole brain pattern, they might have differed.

Validation results (LOEO analyses)
The key output from the LOEO analysis was related to the robustness per cluster. That is, in what probability percentage a given cluster was   Figure 2) Regarding, "Needing" ST ALE-LOEO meta-analysis revealed three peaks with 100% consistent activations in all experiments, within the bilateral middle insula, that included the caudoventral putamen and the claustrum.

Conjunction and contrasts meta-analyses
Contrasts and conjunction analyses were based on ALE results of the two previous ALE results ("Wanting" ST AND "Needing" ST ) that were compared for statistically significant differences and similarities.   (Table 8 and Figure 4).

DISCUSSION
Our goal was to compare the brain activation patterns related to value that comes from the state of "Wanting" ST from the one from the state of "Needing" ST . Our study was thus about when needing a stimulus influences its processing without the wanting component for that stimulus, and when "Wanting" ST happens without "Needing" ST .
To answer this, we used an ALE neuroimaging meta-analysis, comparing consistent brain activation patterns during processing of stimuli in these two states. We used the perception of a cue predicting a reward for "Wanting" ST , and we used the perception of food stimuli in a hungry state as a model for "Needing" ST . We first carried out separate meta-analyses on "Wanting" ST and on "Needing" ST , then we contrasted and intersected them to identify differences and similarities between each of these states. We show that processing a stimulus in a "Wanting" ST state seems more related to activity within the mesolimbic dopaminergic brain areas, nigrostriatal dopaminergic regions, and striatal regions, while processing a stimulus in a "Needing" ST state seems more related to activity in viscerosensory cortices (e.g., midposterior insula) and caudal-ventral putamen (and to some extent the caudate tail). Both states seemed to share consistent activation in the caudate nucleus (head and body) and anterior insula. Compared to "Needing" ST , "Wanting" ST more consistently activated the mesolimbic dopamine: the VTA and ventral striatum and pallidum, and nigrostriatal dopamine regions (i.e., SN and dorsal striatum). Compared to "Wanting" ST , "Needing" ST more consistently activated the mid-posterior insula and ACC, caudo-ventral putamen and Caudate tail, and hippocampus. In the following paragraphs, we will discuss our overall results (the ones consistently found in all our meta-analyses) and how by identifying the brain areas most implicated for each state ("Wanting" ST vs. "Needing" ST ) can help us understand how we attribute different types of value to stimuli.

Overview of consistent activation patterns for Wanting versus Needing
Overall, our results-from main individual meta-analyses, LOEO analyses, and contrasts-confirm that the activation pattern of "Wanting" ST related value shows consistent activation of VTA, ventral striatum, putamen, pallidum, and anterior insula. Our results are in line with previous human studies using a wide range of methods or approaches (Carter et al., 2009;Knutson et al., 2001Knutson et al., , 2003Krebs et al., 2009;O'Doherty, 2004;O'Doherty et al., 2002;Oldham et al., 2018;Schott et al., 2008;Simon et al., 2015;Wilson et al., 2018). For "Needing" ST , our results show that only the middle insula and to some extent the caudal-ventral putamen are consistently related to "Needing" ST related value. The implication of the insula and dorsal striatum in "Needing" ST is in accordance with some previous literature findings (Goldstone et al., 2009;Siep et al., 2009;van der Laan et al., 2011). However, previous meta-analyses and studies on "Needing" ST had also identified other regions such as OFC, ACC, and amygdala/parahippocampal gyrus (Chen & Zeffiro, 2020;Führer et al., 2008;LaBar et al., 2001;Mohanty et al., 2008). This could be due to the fact that we report here, regions that have been consistently found in all our meta-analyses (main, contrasts, and LOEO) and thus use a more stringent approach than in previous meta-analyses. Indeed, when only looking at results from our main meta-analysis, we also identified regions within the OFC, ACC, and the amygdala. Nevertheless, using a more stringent approach, our results showing consistent activation mainly restricted to the mid-posterior insula make sense as it is often considered as the core viscerosensory cortex because it projects to other visceromotor cortices (anterior insula, OFC, ACC) (Barrett & Simmons, 2015); and dense multimodal sensory interoceptive prediction errors converge within the posterior insula to guide interoception (Gehrlach et al., 2020). Thus, by combining contrasts, individual and LOEO meta-analyses approaches, we were able to show that the core regions for "Needing" ST (in this case hunger) seems to be the mid-posterior insula).

"Wanting" ST is more of an emotion than "Needing" ST
Our conjunction results showing consistent activations within the anterior insula for both "Wanting" ST and "Needing" ST could be related to the fact that this region integrates emotional states and is associated with emotional representation of internal states (Craig, 2010), and increases the significance of external stimuli that are relevant with regard to bodily, affective, and sensory information (Menon & Uddin, 2010;Young & Nusslock, 2016). Moreover, based on the fact that the anterior insula plays an important role in awareness (Craig, 2011), our findings suggest that this common activation could be related to our ability to be aware of our wants and needs. It is important to note that the anterior insula was found in the contrast "Wanting" ST -"Needing" ST , but not in "Needing" ST -"Wanting" ST . "Wanting" ST , viewed as reward seeking, is often considered as an emotional state (Panksepp, 2004) that may recruit the anterior insula during reward anticipation  (Knutson et al., 2001;see Craig, 2010). Thus, "Wanting" ST can be thought of as a form of emotional reaction triggered by a Pavlovian cue that predicts a reward. Whereas "Needing" ST (physiological in this case) is usually considered a homeostatic emotion or sometimes a simple sensation, because physiological needs such as hunger do not seem to meet the criteria to be classified as emotions (see Panksepp, 2004), even though they can be seen as homeostatic emotions (Craig, 2003). In light of this, we speculate that, in humans, "Wanting" ST can have more emotional power than "Needing" ST because of the more consistent recruitment of the anterior insula. Though the anterior insula might contribute to turning "Wanting" ST into a more conscious desire/craving (Garavan, 2010;Naqvi et al., 2014), "Wanting" ST can also influence behavior without explicit awareness (Berridge & Robinson, 2003;Strack & Deutsch, 2004;Wei et al., 2017).

Short-term value for "Wanting" ST versus long-term value for "Needing" ST
While consistent activations were found for both states within the striatum, each seems to recruit a different sub-region with the ventral and rostral parts, that is, NAcc, ventromedial caudate and rostroventral putamen more consistenly found for "Wanting" ST and the caudoventral part of the putamen (that often included the tail of the caudate) more consistently found for "Needing" ST . This spatial difference could be related to the functional roles of these subregions including the coding of short-versus long-term values of stimuli. Indeed, our results for "Wanting" ST are in line with findings that suggest that ventral striatum is more responsive to reward or its prediction than the dorsal striatum (Schultz et al., 2000) and that rostral striatum, mainly the caudate head, encodes short term or flexible value (Kim & Hikosaka, 2015). This is also in line with the view of "Wanting" ST as a moment to moment modulation of a cue that predicts reward in synergy with dopaminergic states (Zhang et al., 2009). In contrast, "Needing" ST , was more associated with consistent activation within the caudo-ventral putamen (called "putamen tail", see Kunimatsu et al., 2019) and (to a lesser extent) the caudate tail, both referred to as striatum tail (Amita et al., 2019), regions that acquire long-term values of stimuli based on the historical experience of reward, but not on prediction of rewards (Kunimatsu et al., 2019). Thus, in line with theories and previous studies (Amita et al., 2019;Kim & Hikosaka, 2015;Kunimatsu et al., 2019;Zhang et al., 2009), our results might be interpreted as showing that value representation of a wanted versus needed stimuli rely on distinct regions of the striatum and that this difference could be driven by the temporal aspects or requirement of value processing for each state.

Directional and activational effect of value
The value assigned to stimuli can have a directional effect or activational effect. The directional effect is linked to choice (preference or action selection) and directs towards or away from stimuli, while the activational effect is related to action initiation, maintenance, and vigor (see Salamone et al., 2018). "Wanting" ST AND "Needing" ST metaanalytic conjunction showed that both states consistently activate the caudate nucleus (head and body) and anterior insula (discussed above), regions implicated in action selection (Hollon et al., 2014;Ito & Doya, 2015;Petzschner et al., 2021) and emotional representation of internal states (Craig, 2010). The caudate is involved in goal-directed behavior (Balleine & O'Doherty, 2010;Knutson & Cooper, 2005), and in the pairing between an action and the value of its consequence (Schwabe & Wolf, 2010), such as on the current state of the organism (see Balleine, 1992). Thus, the caudate is implicated in choice/action selectionrelated value (Hollon et al., 2014;Ito & Doya, 2015), and is involved in directional value (Salamone et al., 2016). The implication of the caudate in "Wanting" ST AND "Needing" ST conjunction suggests both states can influence choice/action selection, that is, directional value of stimuli. Thus, by doing a meta-analytic conjunction of "Wanting" ST AND "Needing" ST , we were able to show that both "Wanting" ST and "Needing" ST can influence the directional value of stimuli. However, as we will see, each state seems to rely on distinct neural substrates to compute this directional value.
Directional value for "Wanting" ST seems to arise from activity within the dopaminergic system. The VTA and SN, which contain the main dopaminergic neurons, were shown to be more consistently activated for "Wanting" ST than "Needing" ST . Our results also show that the regions of the ventral striatum, that is, the NAcc and the ventromedial caudate and rostroventral putamen (Haber & Knutson, 2010) were more consistently activated for "Wanting" ST -"Needing" ST , as well as the globus pallidus and the ventral pallidum (VP) (not shown). Indeed, incentive salience "Wanting" ST is generated when a reward cue is synergistically mixed with the state of mesocorticolimbic circuits (which mainly implicates the VTA, NAcc, and pallidum) (Warlow & Berridge, 2021;Zhang et al., 2009). Based on our results, we suggest that the directional value of "Wanting" ST towards stimuli comes from the cortico-striato-midbrain pathway, and first starts with the VTA which computes the prediction error that signals change in expected reward prediction (Schultz et al., 1997) and project mesolimbic dopamine to the ventral striatum (NAcc and VP) (Haber & Knutson, 2010). Second, the activity of the NAcc shell which corresponds to ventrolateral putamen in humans is the final path to the directional value of "Wanting" ST (Holmes et al., 2010), and it is known that mesolimbic dopamine activation within the NAcc or ventral striatum has strong influence on the dorsal striatum (Tricomi et al., 2009).
As mentioned before, with regard to "Needing" ST (the reaction to a needed stimulus), the middle insula, which was found as peak in all of our analyses including our contrasts in favour of "Needing" ST , seems to be the core regions for "Needing" ST (or in this case hunger: when one perceives food while hungry). In this sense, our results confirm that within the insula, it is the middle insula that pairs internal states to relevant external stimuli as argued by Craig (2010). Moreover, our findings dovetail those in the literature that show that the insula plays a role in an "as-if" representation of the bodily state (Damasio, 1994;Naqvi & Bechara, 2010), and that the insula encodes the incentive value of outcomes as a form of incentive memory (Balleine & Dickinson, 2000). Indeed, when hungry or thirsty, the midposterior insula simulates future satiety state in the presence of food or water cues for both humans and animals (Chen et al., 2016;Livneh et al., 2020). Those cues create an interoceptive "prediction error" (see Barrett & Simmons, 2015). Based on interoceptive prediction error from mid-posterior insula, the visceromotor cortices (ACC, OFC, and anterior insula) make predictions about desired internal states (Barrett & Simmons, 2015), and enhance the value of stimuli and actions that fulfill the predictions (Petzschner et al., 2021). Based on our results, we suggest that the mid-posterior insula prediction error might be the origin of the directional value of "Needing" ST in the same logic the VTA does for "Wanting" ST , that is, by computing a sort of prediction error that influences cue selection (see Arsenault et al., 2014); and in this case (i.e., for "Needing" ST ), it is an interoceptive prediction error (Barrett & Simmons, 2015). In this regard, the directional value of "Wanting" ST and that of "Needing" ST depends on two different prediction errors: for "Wanting" ST , the prediction error is computed within the VTA, and for "Needing" ST , the (interoceptive) prediction error is computed within the mid-posterior insula. Importantly, although we focused our meta-analysis on the hunger state and the processing of food stimuli, we think that our results can be generalized to other types of needing states and stimuli. Indeed, it is known that the mid-posterior insula receives multimodal sensory interoceptive signals to compute a prediction error (Livneh et al., 2020).
If both states can give rise to directional/action selection value (albeit differently), only "Wanting" ST seems associated to activiational value. Indeed, consistent activations within the NAcc was only found in our "Wanting" ST meta-analysis, "Wanting" ST -"Needing" ST metaanalytic contrast, and even when stringent LOEO analyses were used.
"Wanting" ST has more (compared to "Needing" ST ) control on activational value because the prediction error signal is sent to the NAcc, which (makes those predictions and) has strong influence to the pallidum which has a lot of impact on invigoration of motor action possibly through a more direct connection to the thalamus (Balleine & O'Doherty, 2010;Haber & Knutson, 2010). In line with literature, our results point out that the activational aspects of cue-induced "Wanting" ST is likely mediated by the mesolimbic dopamine that implicates activation within the central NAcc (see Holmes et al., 2010;Salamone & Correa, 2002;Salamone et al., 1997Salamone et al., , 2016Salamone et al., , 2018 or ventral striatal regions in general (Haber & Knutson, 2010), which have strong influence on the dorsal striatum (Tricomi et al., 2009). Our results that "Needing" ST did not consistently activate dopaminergic regions, are in line with the now admitted fact that needs by themselves do not have activational value (see Salamone et al., 2018) and are not the main source of motivated behavior (Bindra, 1974;Berridge, 2004), although they can amplify it (Toates, 1994). The fact that "Needing" ST has only the directional part (choice/preference or action selection), not the activational one, means that a needed stimulus must still become "wanted," by altering mesolimbic dopamine reactivity and encountering a relevant reward predicting cue (Zhang et al., 2009), in order to have full motivational value (Berridge, 2004;Bindra, 1974;Toates, 1994). Thus, motivation is better explained by incentive salience "Wanting" ST than by "Needing" ST (Berridge, 2004;Bindra, 1974). Nevertheless, "Needing" ST can affect "liking" (see Berridge, 2009) (whether for hunger and food or thirst and water) (Balleine, 1992;Dayan & Balleine, 2002) and can create expectation of "liking" through "cognitive desire" (see Berridge, 2012) towards a needed stimulus. This latter is more goal oriented, and based on declarative memories and on cognitive expectations of act-outcome relations (Berridge, 2012). Thus, "Needing" ST generates cognitive desire, but not necessarily "Wanting" ST (incentive salience) (Berridge, 2012 We think that such exploratory behavior, in absence of any reward cue, is related to what has been called the "seeking system" by Panksepp (2004). This system is composed of the hypothalamus, the ventral striatum and the VTA (Panksepp, 2004), and can include the ventromedial prefrontal cortex (Di Domenico & Ryan, 2017;Panksepp & Biven, 2012), and is responsible for energized exploratory and search behaviors and investigation, and it does not need to be stimulated by a positive incentive cue (Harmon-Jones et al., 2013;Panksepp, 2004). However, when there is a reward cue, activity of that system contributes to adding incentive salience or "Wanting" ST to that cue (Berridge, 2004) and thus turn the exploratory behavior into needinduced "Wanting" ST (see Anselme, 2015). Moreover, it should also be noted that need states can direct and somehow elicit behavior even in absence of the energizing effect of incentive salience "Wanting" ST (Balleine, 2005;Niv et al., 2006;Salamone et al., 2018), and areas of the ventromedial prefrontal cortex and ACC can elicit goal directed behavior based on internal states, even in absence of a reward cue that prompts and guides the animal (Passingham & Wise, 2012).

Incentive cue ("Wanting" ST ) versus outcome relevant cue ("Needing" ST )
Although "Wanting" ST and "Needing" ST are constructs that surely extend beyond the kind of reward, the interpretation of the present results should be done with caution as some brain areas might be indeed influenced by the type of reward, namely the difference between primary (food for "Needing" ST experiments) and secondary (money or points for "Wanting" ST experiments) reward. Of note, "Wanting" ST studies did not explicitly exclude food-related studies and were not limited to secondary rewards, but our selection criteria resulted in the fact that we did not find food-related "Wanting" ST studies to include. However, we believe that, in our study, the difference is not in terms of the kind of reward, but in terms of situations or paradigms, as was also argued by Sescousse et al. (2013) who conducted a meta-analysis between primary and secondary reward, and found higher dopaminergic striatal activation for secondary reward.
They argued that it "is unlikely to be related to the very nature of monetary rewards," but more about the "protocols used" (Sescousse et al., 2013). In other words, it is unlikely that there is more dopaminergic activation within the ventral striatum for monetary (secondary) reward compared to food or sex (primary) reward, and the difference is indeed in the paradigm (situation) used rather than the type of reward (Sescousse et al., 2013). In our view, the cues in the "Wanting" ST experiments were incentive stimuli, that is, stimuli that motivate action and bias behavior (see Robinson et al., 2014); whereas the reward cues included in the "Needing" ST experiments were not, though they were relevant for the current (deprivational) state (see Balleine, 2005). Thus, though the type of reward can influence some brain activation, it is the role (rather than nature) of the reward cue, based on paradigms or situation, that mainly differentiate between "Wanting" ST and "Needing" ST , and generate either mesolimbic-related reward prediction/prediction error or interoceptive prediction/prediction error, respectively. In a similar paradigm (i.e., where a food cue triggers action to gain it), hunger and/or food (used here for "Needing" ST ) would also activate "Wanting" ST , and would result in dopamine-related ventral striatum activation (Simon et al., 2015(Simon et al., , 2016Yousuf et al., 2018). Of note, "Wanting" ST studies did not explicitly exclude food-related studies and were not limited to secondary rewards. In "Wanting" ST experiments that include food, food cues elicited higher mesolimbic dopaminergic response in the ventral striatum when participants were hungry, and such activity was reduced for food cues when participants were satiated (Yousuf et al., 2018). This can be viewed as hunger elevating mesolimbic dopamine, which peaked at the event of the food cue as argued in incentive salience theory (Berridge, 1996(Berridge, , 2004. Thus, "Wanting" ST is closely related to the presence of a cue, associated to either primary or secondary reward, that triggers action or approach behavior based on mesolimbic dopamine state, and that can happen with or without need state (see DiFeliceantonio & Berridge, 2012).
Though we found no study in which participants are in need of money, it is possible that passively viewing money pictures would not necessarily lead to significant mesolimbic dopamine activity, that is, "Wanting" ST , if those pictures are not cues to pursue monetary rewards. Also, it is not sure whether, if one needs money, passively viewing money cues would trigger activity in the mid-posterior insula in the same way as food/hunger or water/thirst (Livneh et al., 2020). Both these hypotheses would need to be empirically tested. In brief, "Wanting" ST and "Needing" ST provide two different roles/values to reward cues, depending on situations or paradigms. "Wanting" ST related cues are stimuli that motivate action and bias behavior (see Robinson et al., 2014), whereas the "Needing" ST related cues are stimuli that are outcome relevant for the current need state (see Balleine, 2005), and those two roles/values do not necessarily (or always) go together. In that sense, our findings and discussion are more about a difference between such cue roles/values, rather than a difference between primary and secondary reward.

Implication on addiction and other maladaptive behaviors
The conceptualization of "Needing" ST , when it happens without "Wanting" ST somehow dovetails recent neurobehavioral theories of addiction (including food addiction, but see Gearhardt et al., 2011) (for a review see Bickel et al., 2018). Our study, showing VTA and ventral striatum for "Wanting" ST , mid-posterior insula for "Needing" ST , and anterior insula for both, allow us to speculate that there might be at least two sources of origin of addiction-like behavior. One would be related to interoception, likely gated by the insula, with the midposterior insula receiving aversive states information (Gehrlach et al., 2020;Livneh et al., 2020) as found in our "Needing" ST -"Wanting" ST contrast; and anterior insula turning such information into emotion, desire and craving (Craig, 2009;Noel et al., 2013;Turel & Bechara, 2016) as shown in our "Wanting" ST -"Needing" ST contrast and conjunction map. Whereas the other source of addictive behavior would be related to dopamine activity, mainly in the striatum (Berridge & Robinson, 2016;Hogarth et al., 2013;Robinson & Berridge, 1993;Volkow et al., 1999) as found in our "Wanting" ST -"Needing" ST contrast. This could lead to difference between "addiction" and "dependence" (see O'Brien et al., 2006), with addiction being related to excessive consumption, and thus more related to "Wanting" ST . Whereas dependence would be more related to aversive states caused by withdrawal symptoms (that can happen in absence of addiction, see O'Brien et al., 2006), and thus more closer to "Needing" ST . The difference could also be temporal, with the beginning of addiction and establishment of sensitization related to striatal dopaminergic areas, thus "Wanting" ST ; and withdrawal symptoms due to aversives states related to perception of aversive internal states and urge to terminate them within the insula, which includes "Needing" ST . Moreover, dopamine-related sensitization theory often separates "liking" from "wanting," where the latter is more responsible for addictive behavior (Berridge & Robinson, 2016).
It might be possible to add "Needing" ST as a separate component. In that case "liking" ST can serve as the beginning of the addictive process, but soon then sensitization happens and is mediated by dopaminerelated areas and "Wanting" ST , and withdrawal is mediated by insula and "Needing" ST .

Limits
We would like to point out some limits to our work. First, "Wanting" ST experiments did not include physiologically related stimuli such as food or water. We argue that the experiments for "Wanting" ST really expressed "wanting" (see Berridge, 2004) because the contrasts we used focused on the processing of the cues, not the outcome, and the cues in those experiments triggered the decision for reward seeking. Indeed, "wanting" has been related to decision utility, that is, the "choice to pursue or consume an outcome" (Berridge & O'Doherty, 2014), induced by a cue . So, although we could not know the mesolimbic state of participants in those experiments, the behavioral situations, that is, a cue that triggered a decision to seek reward, do seem to induce "wanting" (see Berridge, 2004).
Furthermore, "Wanting" ST or motivation activates a general system, regardless of the type of stimulus (Bouton, 2016). Thus, though most stimuli for "Wanting" ST were money or points, the mechanism for cue induced decision for reward seeking, that is, "Wanting" ST is the same.
Second, "Needing" ST contrasts only focused on hunger and the processing of food (physiologically related stimuli). Though our method could seem dependent on the type of physiological state, that is, hunger, and thus the type of needed stimuli, that is, food; our findings and interpretation seem to go in the same direction than studies and theories that suggest an integrative role of all physiological states within the insula. In that sense, Gehrlach et al.'s (2020) results suggest that dense multimodal sensory prediction errors converge in the posterior insula to guide interoception. Based on that, it has been argued that that the insula represents physiological state in cue-independent spontaneous activity, which is then modified directionally by cues that predict water or food availability (Namboodiri & Stuber, 2020). This suggests that our findings regarding the activation of mid-posterior insula might not be hunger/food dependent but rather be resulting from a more general role of the insula with regard to bodily states, specifically serving interoceptive inference (Allen, 2020). Though there are justifications for our methodology in general, it would still be interesting for future studies to test the difference between "Wanting" ST and "Needing" ST with more control on the dopaminergic state, on the type of physiological states and on the type of stimuli.
Nevertheless, wanting and/or needing go way beyond eating, drinking or winning money or points. The distinction between wanting versus needing can apply to virtually any decision, and can even be viewed as philosophical or phenomenological concepts. In that sense, they go way beyond our study and the tools we have used. Thus, our study should be viewed as testing some manifestation of wanting and/or needing rather than testing the general phenomena.

CONCLUSION
Our goal was to compare the brain representation of "Wanting" ST and "Needing" ST related values -two states that guide value attribution and our consumption behaviors. Our results suggest distinct brain systems for both states, with the mesolimbic dopaminergic circuitry as the core for "Wanting" ST and the posterior-middle insula for "Needing" ST .
Whereas "Needing" ST only provides directional value (through an interoceptive prediction error), "Wanting" ST which involves dopamine provides both directional (through a reward prediction error) and activational value. Because "Needing" ST does not provide activational value to stimuli, full motivation (directional and activational) to consume depends more on "Wanting" ST than on "Needing" ST , and means that "Wanting" ST has more power to activate behavior. This might explain why we consume what we want beyond what we need (Stern, 2000).

ACKNOWLEDGMENT
The research was supported in part by NSERC Discovery Grant #RGPIN-2018-05698, UdeM institutional funds, and Mitacs Grant #IT20458.

CONFLICT OF INTEREST
The authors declare that they have no conflict of interest.

AUTHOR CONTRIBUTIONS
Juvenal Bosulu: Designed the study, performed the database search, performed data analysis, interpretation, and wrote the manuscript.

DATA AVAILABILITY STATEMENT
All data are available upon request.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1002/brb3.2713