You are on page 1of 6

Physiology & Behavior 182 (2017) 101–106

Contents lists available at ScienceDirect

Physiology & Behavior

journal homepage:

Explicit wanting and liking for palatable snacks are differentially affected by MARK
change in physiological state, and differentially related to salivation and
Richard J. Stevenson⁎, Heather M. Francis, Tuki Attuquayefio, Candice Ockert
Department of Psychology, Macquarie University, Sydney, Australia


Keywords: Incentive salience theory (IST) suggests that ‘wanting’ and liking are dissociable processes. We argue that explicit
Liking measures of wanting in humans can reflect the impact of implicit ‘wanting’ as envisaged by IST, suggesting that
Wanting dissociations should also be evident for explicit judgments of wanting and liking. To test this, participants were
Palatability asked to make ratings of these variables for 8 palatable snack foods – and in a related test salivation rate was also
assessed. Participants viewed and sniffed each snack food and rated wanting, and then sampled it and rated
Ingestive behavior
liking and whether they wanted more of it. Following a lunch eaten to satiety, and composed in part of half of the
palatable snack foods, participants repeated their evaluations of the snack foods (and salivation rate). Liking
changed less across lunch than wanting and want more ratings, the last-mentioned changing the most. Change in
liking was associated with change in salivation rate, independent of wanting, and change in wanting was as-
sociated with change in hunger independent of liking. We argue these dissociations are consistent with ‘wanting’
influencing explicit wanting, and that want more ratings may represent a ‘purer’ measure of IST ‘wanting’.

1. Introduction and reward structures. The complications ensue for IST because humans
are capable of explicitly reporting on states that seem to parallel those
The motivational basis of ingestive behavior is important to un- that govern a rat's behavior. Thus, when a food deprived rat bar presses
derstand as its dysfunction may contribute to excess weight gain and more frequently to earn chow, relative to a sated rat, it is easy to see
disordered eating (e.g., [11,24,33]). An important account of motiva- this in human terms using concepts such as wanting and desire –terms
tion, developed in animals, is incentive salience theory (IST; [2,28]). that are widely used in the human eating literature (e.g., see [34]).
Two processes are central to IST. The first is wanting (sometimes re- However, there are significant concerns over such parallels between
ferred to as desire to eat and more distantly perhaps, hedonic hunger), particular human states of mind (i.e., wanting food) and animal beha-
namely the drive to obtain biologically significant goals that occurs vior in the laboratory [36]. As the authors of IST note, wanting in an-
when a learned cue associated with that goal is perceived in the pre- imals is preconscious and on this basis, they claim that the parallel state
sence of an appropriate physiological state. The second is liking, namely in humans is also preconscious [3]. Consequently, any similarity be-
the reward process that ensues when the goal object is consumed. While tween wanting in animals and explicit mental states in humans may be
translating the IST notion of liking from animals to humans has gen- incorrect. This has led to the study of two forms of wanting in humans –
erally been straightforward, the translation of wanting has been more implicit ‘wanting’ (often presented in single quotes, as here, to indicate
problematic as outlined below (e.g., see [26,36]). Complications have the difference from colloquial usage), which arguably parallels the
also arisen in attempts to show dissociations between wanting and preconscious form studied in rats, and explicit wanting, which refers to
liking in humans (e.g., [15,16]), even though these have been demon- mental states such as wanting and desire. This distinction has spawned
strated in animals (e.g., [38]) and are seemingly evident in humans two forms of measurement, one involving implicit measures to tap
(e.g., wanting something that no longer provides reward). Here we ‘wanting’ (e.g., [13]) and the other self-reports to tap explicit wanting
demonstrate a novel, simple and powerful means of dissociating (e.g., [4]). Our focus here is with dissociating explicit wanting from
wanting and liking in the context of food. liking - as explicit measures dominate the human IST literature and are
Humans and rats share much of the same basic neural motivational easy to assess [26].

Corresponding author at: Department of Psychology, Macquarie University, Sydney, NSW2109, Australia.
E-mail address: (R.J. Stevenson).
Received 10 February 2017; Received in revised form 7 September 2017; Accepted 6 October 2017
Available online 13 October 2017
0031-9384/ © 2017 Elsevier Inc. All rights reserved.
R.J. Stevenson et al. Physiology & Behavior 182 (2017) 101–106

IST defines ‘wanting’ as state dependent [2,28]. Thus, viewing a cue changes in hunger should track changes in physiological state.
to food in different states (e.g., hungry vs. sated) should generate sig-
nificant shifts in ‘wanting’ and presumably the same applies to explicit 2. Method
wanting too. Hedonic reactions to food also change with alterations in
physiological state - alliesthesia [6] – as well as more specifically for 2.1. Design overview
foods that have just been eaten relative to those that have not – sensory
specific satiety [30]. The dissociation we wish to test is that liking will The study used a fully-within subject design. Participants viewed
change to a significantly smaller degree than explicit wanting, fol- and sniffed palatable snack foods (both sweet and savoury, and in-
lowing a relevant change in physiological state (i.e., energy depletion to cluding items to be eaten at lunch and not eaten at lunch) and made
repletion). ratings of how much they wanted to consume them now. A morsel of
There are two reasons for thinking that liking will change less with each snack food was then consumed followed by ratings of liking and
state than wanting. First, theoretically, wanting for food should be wanting more. Salivation rate to the same foods was also measured in a
terminated by the change from energy depletion to repletion, thus al- separate test. All participants then consumed lunch, after which the
lowing a shift in goals towards other targets (e.g., satisfying thirst). Any same evaluations were repeated for the same palatable snack foods, as
concurrent change in hedonic responsiveness should be less dramatic well as a second salivation rate test.
(i.e., alliesthesia and sensory-specific satiety – more below), because
large shifts in liking across a meal would be maladaptive. This is be- 2.2. Participants
cause they could later adversely affect food choice, as recalling end of
meal ‘disliking’ could impair consumption for a food that was actually Thirty-six participants, 18 male, took part for course credit. Mean
safe, nutritious and initially palatable. Second, empirically, there is age was 19.3 years (range 18–25) and participants ranged from under-
already some evidence favouring a dissociation between ‘wanting’ and weight to obese (range 16.6 to 31.6), with 25 falling in the normal
liking after a change in state. This has been observed using measures of range (overall sample M = 21.9). One participant reported currently
implicit ‘wanting’ and explicit liking [12], but differences solely based taking stimulant medication for ADHD and this person's data was ex-
on explicit wanting and liking are also evident. Finlayson et al., (2008) cluded from the analysis leaving 35 cases. All participants provided
reported ratings of wanting and liking at the start and end of a pizza written informed consent and the protocol was approved by the
meal, and while no change measure was statistically tested (i.e., change Macquarie University Human Research Ethics Committee.
in liking over lunch vs. change in wanting over lunch), the means
suggest a much greater change in wanting for the pizza relative to liking 2.3. Materials
for it.
The closest to a direct test of our current hypothesis was reported by The evaluation test stimuli consisted of: (1) one cheddar cheese
Attuquafeio et al., [1]. They measured explicit wanting and liking for a cube (Mainland, approximately 1cm3); (2) one cheese and bacon ball
set of snack foods prior to and after a filling lunch. The study found that (Fritolay); (3) one BBQ flavoured Pringle; (4) one salt and vinegar
the decline in explicit wanting across lunch was larger than the decline flavoured Pringle; (5) one piece of caramel fudge (Woolworths, ap-
in liking, and that the magnitude of this difference was moderated by proximately 1cm3); (6) one TimeTam bite (a chocolate covered mini-
how frequently the participant consumed foods rich in saturated fat and biscuit); (7) one Malteser (Mars); and (8) one vanilla wafer mini-biscuit
added sugar - this being the main focus of the study. While these results (Woolworths).
are encouraging, they are also problematic. This is because lunch in- Lunch consisted of three courses. The starter, presented in two
cluded some of the same items that were also present in the snack foods bowls, was composed of two savoury snack foods selected (counter-
evaluated before and after the meal. This may have acted to reduce the balanced across participants) from the four used in the evaluation test.
degree of change in liking for the snack foods, relative to changes in The amounts were selected so as to be broadly equivalent in volume
explicit wanting, because of dilution of sensory specific satiety (i.e., (34 g of cheese cubes [578KJ]; 10 g of cheese and bacon balls [230KJ];
occurring for some but not all snack foods). If both sensory specific 17 g of BBQ Pringles [374KJ]; 17 g of salt and vinegar Pringles
satiety and alliesthesia effects had occurred, then changes in liking for [374KJ]). Main course was either beef (1340KJ) or vegetarian
the snack foods may have been far greater, thus potentially equalling (1140KJ) lasagne (Woolworths; both 260 g). Desert, presented in two
changes in explicit wanting. A better test would be to determine the bowls, involved two sweet snack foods selected (counterbalanced
impact on liking of alliesthesia and sensory specific satiety, and then see across participants) from the four used in the evaluation test. The
if changes in explicit wanting still exceed these effects alone and to- amounts were again selected so as to be broadly equivalent in volume
gether. (34 g of caramel fudge [544KJ]; 34 g of Tim Tam bites [782KJ]; 31 g of
The design of the current study was based upon Attuquafeio et al., Maltesers [713KJ]; 31 g of vanilla wafer mini-biscuits [589KJ]).
[1]. As briefly described above, it involved evaluating palatable snack
foods (i.e., assessing wanting on sight/smell, and liking and wanting 2.4. Procedure
more after tasting), eating a meal to repletion and then re-evaluating
the same snack foods again. A key design change was to specifically All participants were asked to breakfast as per normal, and then
determine changes in liking from alliesthesia alone (i.e., for snack foods refrain from eating in the 3 h before testing so as to arrive hungry for
not eaten at lunch) from those for alliesthesia and sensory specific sa- lunch (i.e., arriving between 11 am and 1.30 pm). Participants were
tiety combined (i.e., for snack foods eaten at lunch). This would ensure told that they could drink water in this period but not caloric beverages.
more careful control of state-dependent changes in liking. In addition, All participants reported complying with these instructions.
we also tested for other evidence of a dissociation between explicit Participants were asked to rate how hungry, full, thirsty, happy, sad,
wanting and liking. First, we measured salivation rate towards the alert and relaxed they were on 120 mm line rating scales (anchors Not
palatable snack foods, both before and after lunch, to establish whether at all to Very) – hereafter termed the rating set (note that only hunger
changes in salivation rate were better explained by changes in liking for and fullness data are described here as the remaining ratings did not
the same snack food or by changes in explicit wanting. This particular reveal anything of interest). The first part of the pre-lunch evaluation
comparison has not been made before, although salivation rate has test then commenced. Participants had two pieces of pre-weighed
been shown (albeit inconsistently) to be linked to palatability [23,37] dental wadding placed around their submaxillary and sublingual ducts
and to hunger too [23,37]. Second, we also examined whether changes (i.e., one piece on either side of the floor of the mouth under the
in hunger were related to changes in explicit wanting, on the basis that tongue), as well as a further two pieces around the parotid ducts (i.e.,

R.J. Stevenson et al. Physiology & Behavior 182 (2017) 101–106

one piece each side of the upper jaw, between gum and cheek). The (State - pre-lunch vs. post-lunch; Evaluation type - Wanting vs. Liking vs.
length of time (in secs) this wadding was in place (from the last piece Want more; Exposure - Snack foods eaten at lunch vs. Not eaten at
going in to the last piece coming out) was recorded for each participant. lunch; Taste - Sweet vs. Savoury snacks), two further variables were
The participant was then asked to go through a specific sequence of included in the analysis. The first was gender, as this might affect he-
actions for each of the 8 snack foods, which were presented in rando- donic responses to the snacks for several reasons (e.g., greater re-
mised order. For each piece of snack food, the experimenter instructed straint). The second, which was included as a covariate, was the
the participant to look at it twice, sniff it twice and touch it twice, amount of energy consumed at lunch (following Z-transformation).
before moving on to the next item. After all 8 food items had been These data and their residuals were normally distributed and so they
examined in this way the wadding was removed for later weighing, and were analysed using a mixed-design ANCOVA (SPSS 24, GLM com-
participants were asked to rinse their mouths with water. mand).
The second of the pre-lunch evaluation tests then commenced. Participants saliva was collected over the whole period of the first
Participants were presented with a new set of the same eight snack pre-lunch evaluation test, which lasted 2 mins and 26 s (SD = 44.6 s),
foods in a different random order. Starting with the first item, partici- and over the whole period of the first post-lunch evaluation test, which
pants were asked to look at it carefully twice and then judge how much lasted 2 mins and 2 s (SD = 34.6 s). Pre- and post salivation rate data
they wanted to eat it using a 120 mm line rating scale (anchors Not at (i.e., grams of saliva produced per minute) and their difference score
all and A lot). They were then asked to consume the sample after which (Post - Pre) were not suitable for parametric analysis as these data were
they made two further ratings, both again on 120 mm line rating scales: skewed and kurtotic (all Z's > 3). Consequently, these data were
(1) How much do you like this food? (anchors Not at all and A lot); and analysed non-parametrically.
(2) How much more of this food would you like to eat now? (anchors Finally, to establish whether changes (all involving Post - Pre sub-
None and A lot). Following a water rinse, participants then repeated traction) in liking, wanting and wanting more on the evaluation test
this process for each of the remaining snack food items. were differentially related to changes in salivation rate, hunger and
Before starting lunch, participants completed their second rating set fullness across lunch, a series of correlational analyses were performed.
(i.e., hunger, fullness etc). Participants were then instructed to eat until For the tests involving changes in salivation rate Spearman's rho was
they were comfortably full and told that all uneaten food would be used, with Pearson's r used for the remaining variables. Partial corre-
thrown away. Ad libitum chilled water was available throughout lunch. lations (Spearman or Pearson) were then calculated to see if any
First, they were served the starter, which consisted of two bowls each identified relationship with either wanting or liking was still significant
containing a different savoury snack (see Materials). Five minutes was after controlling for the other variable (i.e., liking or wanting, respec-
provided to consume this food. They were then given 10 mins to con- tively).
sume their choice of beef or vegetarian lasagne. This was then followed
by desert, which consisted of two bowls each containing a different 3. Results
sweet snack (see Materials), again with 5 mins available for this phase.
Participants were allowed to read magazines (content screened to avoid 3.1. Lunch
any eating-related or upsetting material) while they consumed lunch.
All uneaten food was removed for later weighing. To confirm that participants underwent a change in state, we ex-
Immediately after completing a third rating set (i.e., hunger, full- amined both their energy intake data, and self-reports of hunger and
ness etc), participants undertook a repeat of the two-part evaluation fullness, across the meal.
test. This was conducted in an identical manner to the first evaluation All participants consumed lunch, with a mean overall intake of
test, starting with the salivation rate component and then followed by 2404KJ (SD = 594; range = 1445–3371). This was composed of a
the evaluation part - as described in detail above. Presentation orders mean intake of 599KJ for the savoury starter (SD = 218;
for the snack food items on both tests was identical to that used for each range = 183–978), a mean intake of 998KJ for the lasagne (SD = 247;
participant on the pre-lunch test. After completing both tests and a final range = 400–1286) and a mean intake of 806KJ for the sweet dessert
rating set, participants were asked to fill in a brief biographical measure (SD = 334; range = 61–1500).
(including medical-related dietary issues; eating disorders, diabetes, Hunger ratings remained stable from the first (M = 69.4;
medications that can affect appetite etc). All participants were then SD = 21.8) to the second rating (M = 63.6; SD = 27.6), but decreased
weighed and had their heights measured - both without shoes. significantly after lunch (M = 11.8; SD = 13.7; 2nd vs. 3rd rating,
paired sample t-test, t(34) = 11.47, p < 0.013), then remained stable
2.5. Analysis to the end of the experiment (M = 10.3; SD = 15.0). Fullness ratings
increased from the first (M = 29.5; SD = 20.9) to the second rating
Hunger and fullness data were analysed using t-tests (correcting for (M = 47.0; SD = 24.1; 1st vs. 2nd rating, paired sample t-test, t(34)
multiple comparisons; alpha set at 0.05/4), as we were only interested = 6.08, p < 0.013), and increased further following lunch (M = 91.4;
in specific comparisons so as to determine if the state-related manip- SD = 28.4; 2nd vs. 3rd rating, paired sample t-test, t(34) = 9.73,
ulation (i.e., lunch) had been effective. p < 0.013) and then remained stable to end of the study (M = 102.4;
For the evaluation test, which was conducted twice, once before and SD = 19.9). Thus, participants felt hungrier and less full at the first set
once after lunch, each participant sampled eight snack foods. Of these of evaluation tests before lunch than they did at the second set after
eight foods, four were sweet and four savoury, and of the four sweet, lunch.
two were eaten at lunch and two were not - as with the four savoury
foods, two being eaten at lunch and two not. The assignment of the four 3.2. Evaluation test
sweet and four savoury snack foods to each of these cells of the design
was fully counterbalanced across participants, and for this reason we The analysis of the wanting, liking and want more ratings from the
formed four categories of snack food: (1) sweet and experienced at pre and post-lunch evaluation test were analysed using a five-way
lunch; (2) sweet and not experienced at lunch; (3) savoury and ex- mixed design ANCOVA, with State (pre-lunch vs. post-lunch),
perienced at lunch; and (4) savoury and not experienced at lunch. Each Evaluation type (Wanting vs. Liking vs. Want more), Exposure (Snack
of these snack food categories had three associated ratings (wanting, foods eaten at lunch vs. Not eaten at lunch) and Taste (Sweet vs.
liking, want more), with these ratings obtained before and after lunch. Savoury snacks) as the within factors and Gender as the between factor
Thus, in total, each participant had 24 data points available for the and energy intake at lunch as the covariate. Significant main effects and
evaluation test analysis. In addition to these four within-subject factors interactions are detailed in Table 1.

R.J. Stevenson et al. Physiology & Behavior 182 (2017) 101–106

Table 1
Main effects and interactions from the analysis of the pre- and post-lunch evaluation test

Effect F(dfs) = p< Partial eta-

squared =
State F(1,32) = 164.89 0.001 0.84
State by Evaluation type F(2,64) = 18.43 0.001 0.37
Evaluation type F(2,64) = 71.98 0.001 0.69
State by Exposure F(1,32) = 4.90 0.05 0.13
Evaluation type by Exposure F(2,64) = 6.16 0.005 0.16
State by Taste F(1,32) = 23.36 0.001 0.42

Fig. 2. Scatter plot of change in salivation rate against change in liking, for the palatable
snack foods, with these measures being obtained before and after a filling lunch.

by State (F = 0.31). Finally, there was an interaction between State and

Taste (see Table 1), with all evaluations of the sweet foods falling more
across lunch (M change = −34.2; SD = 18.1) than for savoury foods
(M = − 23.3; SD = 15.1). That is state had a larger impact on the
evaluation of sweet foods than on savoury foods across all ratings.
Participants median salivation rate was 0.46 g/min prior to lunch
and 0.53 g/min after lunch. This difference was significant (Wilcoxon
test; Z = 3.01, p = 0.003) indicating that salivation rate on the post
lunch evaluation test was generally greater than on the pre lunch test.
We then examined whether the change in salivation rate across lunch
was associated with changes in liking, wanting or wanting more ratings
across lunch. Neither changes in wanting (rho = 0.05) nor wanting
more (rho = 0.09) were associated with changes in salivation rate, but
Fig. 1. Mean (and standard error) liking, wanting and want more ratings for a set of
palatable snack foods obtained prior to and after a filling lunch. Changes in all three changes in liking were (rho(35) = 0.36, p = 0.036) – see Fig. 2. We
rating types across lunch were significant, with changes in wanting significantly ex- then examined whether this relationship would still be significant if
ceeding changes in liking, and with changes in wanting more, significantly exceeding changes in wanting and wanting more were partialled out. The corre-
changes in liking and wanting. lation remained significant (rhop(31) = 0.39, p = 0.026). These find-
ings indicate that where liking changed little across lunch, salivation
Not surprisingly, there was a main effect of State (see Table 1), with rate tended to be larger post-lunch, relative to pre-lunch - independent
all ratings reported as lower following lunch (M = 38.0/120; of changes in wanting and wanting more.
SD = 13.2) than before lunch (M = 66.7/120; SD = 12.6). Im- We also examined if change in hunger and fullness across lunch
portantly, this was qualified by an interaction with Evaluation type (see were correlated with changes in wanting, liking, wanting more and
Table 1), which is illustrated in Fig. 1. Here it can be seen that liking salivation rate. There were no associations with changes in fullness
ratings declined substantially less across lunch (M change = − 20.5; across lunch, but changes in hunger were associated with changes in
SD = 15.0) than either wanting ratings (M change = − 29.0; wanting (r(35) = 0.41, p < 0.02 – see Fig. 3) and in wanting more (r
p = 0.003; SD = 18.4) or want more ratings (M change = − 36.7; (35) = 0.49, p < 0.005) but not with liking (p = 0.24) or salivation
p < 0.001; SD = 15.6). The change in want more ratings also ex- rate (p = 0.29). Again, we checked to see if the two associations be-
ceeded that of the wanting ratings (p = 0.004). This suggests that tween changes in wanting (and wanting more) and hunger would still
evaluations of wanting based just upon viewing, smelling and touching be significant after partialling out changes in liking. They were (rp(32)
a palatable snack food, are more sensitive to physiological state than = 0.45, p < 0.01 [wanting]; rp(32) = 0.50, p < 0.005 [wanting
evaluations of liking based upon actually consuming it. more]), indicating that greater reductions in hunger across lunch were
In addition to these findings (and a main effect of Evaluation type), associated with greater reductions in wanting across lunch, in-
three further effects were obtained. First, State interacted with dependent of reported changes in liking.
Exposure (see Table 1), so collapsing across all evaluation types, ratings
fell more for snack foods that were eaten at lunch (M change = − 31.0;
4. Discussion
SD = 15.1) relative to those not eaten at lunch (M change = − 26.4;
SD = 15.2). Second, Exposure also interacted with Evaluation type (see
Differentiating explicit wanting and liking has attracted consider-
Table 1), with little difference between snack foods that had and had
able research attention [26]. For ingestive behavior there are relatively
not been experienced at lunch on liking (M difference = 0.8;
few examples of specific dissociations between explicit wanting and
SD = 14.5) and want more (M difference = 0.2; SD = 13.4) ratings,
liking, these either being in specific populations (obese; [25]) or just
but with a larger difference for wanting (M difference [had minus not
using pictorial representations of food [5]. In the experiment reported
had] = − 4.1; SD = 14.7). That is snack foods not consumed at lunch
here, we aimed to determine if explicit wanting declined more fol-
tended to be wanted more, noting that this interaction was not qualified
lowing a change in physiological state (energy depletion to repletion)

R.J. Stevenson et al. Physiology & Behavior 182 (2017) 101–106

wanting judgments to enable a dissociation to occur from liking under

conditions of a change in physiological state.
Needless to say, this is not the only possible interpretation of our
findings. An alternate viewpoint would be that wanting (and wanting
more) ratings simply reflect another set of processes (see [26]), namely
between participants predictions as to how much they may like a food
now (expected-liking) and how much they actually like it when they eat
it (liking ratings). We suggest this perspective does not offer as com-
pelling an account of the dissociation observed here, relative to IST.
There are two reasons for this assertion. First, expected-liking draws
upon a weighted average of prior liking experiences that is distorted by
start (i.e., primacy), peak, and end (i.e., recency) evaluations [21,39].
Most work on expected liking has been conducted outside of the food
domain, but what there is suggests that something broadly similar
happens here too (e.g., [14,20,29]). As expected-liking relies upon an
averaging, rather than being based upon separate memories of liking for
that food when hungry, and liking for that food when sated, there
should then be little change between evaluations made before and after a
meal. So, if expected-liking primarily drives wanting ratings, wanting
Fig. 3. Scatter plot of change in wanting for the palatable snack foods across lunch ratings should be relatively stable across lunch. Clearly, this is not
against change in hunger across lunch. consistent with what we observed, as wanting ratings changed more
than liking, something that would not then be expected if wanting
than liking, when these evaluations were made for palatable snack ratings were primarily driven by expected-liking.
foods sampled before and after lunch in healthy participants. Irre- There is a second reason to suspect that expected-liking does not
spective of whether the snack foods had or had not formed part of solely account for the behavior of wanting ratings. As noted in the
lunch, changes in liking were smaller than participants reported Results (and see Fig. 1), want more ratings changed to a greater extent
wanting to consume them. As part of the design two explicit wanting across lunch than wanting ratings. Recall that participants made the
measures were obtained, one based on just looking at and smelling the want more ratings just after the liking ratings, and both of these almost
food, with the other made after consuming it. While both wanting immediately after consuming an item of the snack food. If then
ratings evidenced greater change than the liking measure, the one made memory-based expected-liking were being used to drive wanting rat-
after consumption demonstrated the largest change across lunch. Par- ings, then want more ratings should be very similar to liking ratings
ticipants salivation rate to the palatable snack foods was also assessed. (i.e., expected liking should fall into line with actual liking). As the data
Salivation rate increased after lunch, when participants again viewed show this was not the case, and in fact want more ratings demonstrated
and sniffed these foods. Changes in salivation rate across lunch were a more extreme change across lunch than the wanting ratings obtained
linked to changes in liking, independent of explicit wanting. Partici- after looking at and smelling the food. To the extent that expected-
pants who reported smaller changes in liking across lunch were most liking is the principal alternate account of how wanting ratings are
likely to salivate more to the snack foods after than before lunch. formed, it is for these reasons that we suggest that IST may better ex-
Conversely, changes in explicit wanting were linked to changes in plain the observed dissociation between explicit wanting and liking.
hunger (but not fullness) across the meal, with this effect being in- As just described, want more ratings changed most across lunch, and
dependent of changes in liking. These various findings suggest that to a greater extent than wanting ratings. One explanation for this dif-
explicit reports of wanting and liking can be differentiated by a change ference could be the relative influence of expected-liking ratings, which
in physiological state, and that related differences also emerge in their may be greater for wanting ratings than for want more ratings. The
relationships with changes in salivation rate (liking) and hunger main reason for suggesting this comes from findings made in the
(wanting). Attuquaefio et al., (2016) study. They found that individual difference
While it is evident that self-reports of wanting and liking behave in changes in explicit wanting relative to liking, across lunch, were
differently following a change in physiological state, it is quite another associated with a measure of episodic memory. However, this was not
thing to infer from this that these represent dissociations relevant to the case for the difference between want more and liking ratings. The
IST. Our principal argument that they do is that the final common reason why this distinction may be important, is that it has been sug-
pathway that informs explicit wanting ratings draws upon multiple gested that expected liking ratings are a product of episodic memory
sources of information, including implicit ‘wanting’ processes that are processes (e.g., [27]). Thus, we suggest that wanting ratings are a
described in IST. Indeed, if implicit ‘wanting’ is an important driver of combination of IST ‘wanting’ and expected-liking, while the want more
human behavior it must ultimately manifest in explicit mental states ratings may be a purer measure of IST ‘wanting’ that is less con-
(e.g., craving) and/or self-observable behavior (e.g., food seeking). That taminated by expected-liking and thus less related to a measure of
this view is to some extent shared by other researchers can be seen in episodic memory. On this basis then, and to the extent that implicit
the widespread use of explicit wanting measures in the human IST lit- ‘wanting’ measures better tap ‘wanting’ in the IST sense, we would
erature [26] and also by the use of explicit wanting as a form of cri- predict that explicit want more ratings made after tasting a food would
terion validity for implicit ‘wanting’, especially in the drug craving be more strongly linked to implicit ‘wanting’ measures, than wanting
literature (e.g., [17–19]). Moreover, there are now several published ratings made after viewing and smelling a food.
papers, outside of the domain of food (and see [10], for an interesting In addition to explicit wanting and liking, we also measured sali-
food-related case), that indicate dissociations for explicit wanting and vation rate to the palatable snack foods, both before and after lunch.
liking that directly mirror implicit measures of these same variables Salivation rate was, on average, significantly higher after lunch than
(e.g., [7]) and even some cases where explicit dissociations are evident before. Several variables are known to affect salivation rate [8] and two
but not with implicit measures [35]. Returning to our findings, liking as of these may relevant here. First, having recently consumed food prior
conceptualised in IST, is then differentiated from explicit wanting, be- to a salivation measure is known to increase unstimulated saliva flow,
cause sufficient components of implicit ‘wanting’ feed into explicit and this could account for the observed change across lunch [9].
Second, consuming water at lunch, which may have increased

R.J. Stevenson et al. Physiology & Behavior 182 (2017) 101–106

participants degree of hydration relative to before lunch, can also act to 10 (2010) 324–334.
[8] C. Dawes, Physiological factors affecting salivary flow rate, oral sugar clearance,
increase salivation rates [31]. In addition to the general trend for in- and the sensation of dry mouth in man, J. Dent. Res. 66 (1987) 648–653.
creased salivation rate across lunch, it is also apparent that there was [9] C. Dawes, F. Chebib, The influence of previous stimulation and day of the week on
variability in the degree to which this occurred (see Fig. 2). We ob- the concentration of protein and the main electrolytes in human parotid saliva,
Arch. Oral Biol. 17 (1972) 1289–1301.
served that this variability was differentially correlated with changes in [10] M. Dermiki, J. Prescott, L. Sargent, J. Willway, M. Gosney, L. Methven, Novel fla-
liking but not changes in wanting. Smaller reductions in liking across vours paired with glutamate condition increased intake in older adults in the ab-
lunch were associated with larger increases in salivation rate across sence of changes in liking, Appetite 90 (2015) 108–113.
[11] L. Epstein, K. Carr, H. Lin, K. Fletcher, Food reinforcement, energy intake, and
lunch. Yet again, this would suggest that explicit wanting and liking are macronutrient choice, Am. J. Clin. Nutr. 94 (2011) 12–18.
not measuring identical constructs, and that sufficient difference exists [12] G. Finlayson, N. King, J. Blundell, Is it possible to dissociate ‘liking’ from ‘wanting’
between them to obtain dissociations even with a relatively modest for foods in humans? A novel experimental procedure, Physiol. Behav. 90 (2007)
sample size used here.
[13] G. Finlayson, N. King, J. Blundell, The role of implicit wanting in relation to explicit
A measure that we would expect to be conceptually closer to liking and wanting for food: implications for appetite control, Appetite 50 (2008)
‘wanting’, relative to liking, is hunger. A significant complication here 120–127.
is that hunger has multiple meanings [32], covering stomach-related [14] E. Garbinsky, C. Morewedge, B. Shiv, Does liking or wanting determine repeat
consumption delay? Appetite 72 (2014) 59–65.
physical sensations, an awareness of having eaten, and most relevant of [15] R. Havermans, “You say it's liking, I say it's wanting”; on the difficulty of disen-
all ‘hedonic hunger’ the desire to consume palatable foods [22]. Con- tangling food reward in man, Appetite 57 (2011) 286–294.
sistent with this, we observed stronger links between changes in [16] R. Havermans, T. Janssen, J. Giesen, A. Roefs, A. Jansen, Food liking, food wanting
and sensory specific satiety, Appetite 52 (2009) 222–225.
wanting and wanting more ratings and changes in hunger, than for [17] K. Houben, R. Wiers, Are drinkers implicitly positive about drinking alcohol?
liking ratings (and interestingly in light of the discussion above re- Personalizing the alcohol-IAT to reduce negative extrapersonal contamination,
garding the difference between wanting and wanting more ratings, a Alcohol and Alcholism 42 (2007) 301–307.
[18] K. Houben, R. Wiers, Personalizing the alcohol-IAT with individualized stimuli:
somewhat stronger link for the latter). As hunger ratings are likely to relationship with drinking behaviour and drinking-related problems, Addict. Behav.
reflect a composite of inputs, including ‘hedonic hunger’ - a construct 32 (2007) 2852–2864.
that shares resemblance to the IST construct of wanting - we would [19] K. Houben, R. Wiers, Implicitly positive about alcohol? Implicit positive associa-
tions predict drinking behaviour, Addict. Behav. 33 (2008) 979–986.
suggest that it is this overlap that accounts for the greater link to
[20] D. Just, O. Sigirci, B. Wansink, Week-end pizza: prices delay evaluations of quality,
wanting/wanting more ratings than to liking. J. Prod. Brand Manag. 24 (2015) 770–778.
In conclusion, we observed a dissociation between changes in ex- [21] D. Kahneman, J. Riis, Living and Thinking about it: Two Perspectives on Life, in:
F. Huppert, N. Baylis, B. Keverne (Eds.), The Science of Well Being, OUP, Oxford,
plicit wanting, liking and wanting more, for palatable snack foods,
2005, pp. 285–304.
evaluated before and after a filling lunch. We suggest that this dis- [22] M. Lowe, M. Butryn, Hedonic hunger: a new dimension of appetite, Physiol. Behav.
sociation reflects the input of implicit ‘wanting’ processes, as envisaged 91 (2007) 432–439.
by IST, into the explicit wanting ratings, allowing a dissociation with [23] R. Mattes, Nutritional implications of the cephalic phase salivary response, Appetite
34 (2000) 177–183.
liking to emerge under conditions of a change in physiological state. IST [24] D. Mela, Eating for pleasure or just wanting to eat? Reconsidering sensory hedonic
‘wanting’ inputs may be purer for ratings of wanting more obtained responses as a driver of obesity, Appetite 47 (2006) 10–17.
after sampling a food than for wanting ratings made after viewing and [25] C. Ochner, E. Stice, E. Hutchins, L. Afifi, A. Geliebter, J. Hirsch, J. Teixeira, Relation
between changes in neural responsivity and reductions in desire to eat high-calorie
smelling alone. We also observed differential patterns of association, foods following gastric bypass surgery, Neuroscience 209 (2012) 128–135.
with changes in liking linked to changes in salivation across lunch, and [26] E. Pool, V. Sennwald, S. Delplanque, T. Brosch, D. Sander, Measuring wanting and
changes in hunger linked to changes in wanting/wanting more ratings liking from animals to humans: a systematic review, Neurosci. Biobehav. Rev. 63
(2016) 124–142.
across lunch. Finally, the simplicity with which explicit wanting ratings [27] E. Robinson, Relationships between expected, online and remembered enjoyment
can be obtained, their easy use with individual stimuli (foods or pic- for food products, Appetite 74 (2014) 55–60.
tures), suggests that understanding how explicit wanting relates to [28] T. Robinson, K. Berridge, The neural basis of drug craving: an incentive-sensitiza-
tion theory of addiction, Brain Res. Rev. 18 (1993) 247–291.
implicit ‘wanting’ as described in IST, is a question worth pursuing. [29] E. Robinson, J. Blisset, S. Higgs, The influence of recent tasting experience on ex-
pected liking for foods, Food Qual. Prefer. 27 (2013) 101–106.
Acknowledgements [30] B. Rolls, E. Rolls, E. Rowe, K. Sweeney, Sensory specific satiety in man, Physiol.
Behav. 27 (1981) 137–142.
[31] I. Shannon, The Biochemistry of Human Saliva in Health and Disease, in: H. Rowe
We thank the ARC for their support (DP150100105) and Andrea (Ed.), Salivary Glands and their Secretions, University of Michigan Press, Ann
Zuniga, Sophia Pender, Christine Leonards for assistance with this ex- Arbor, 1972, pp. 94–121.
periment. [32] R.J. Stevenson, M. Mahmut, K. Rooney, Individual differences in the interoceptive
states of hunger, fullness and thirst, Appetite 95 (2015) 44–57.
[33] E. Stice, S. Spoor, J. Ng, D. Zald, Relation of obesity to consummatory and antici-
References patory food reward, Physiol. Behav. 97 (2009) 551–560.
[34] R. Stubbs, D. Hughes, A. Johnstone, E. Rowley, C. Reid, M. Elia, R. Stratton,
H. Delargy, N. King, J. Blundell, The use of VAS to assess motivation to eat in
[1] T. Attuquayefio, R.J. Stevenson, R.A. Boakes, M. Oaten, M. Yeomans, M. Mahmut, human subjects: a review of their reliability and validity with an evaluation of new
H.M. Francis, Increased consumption of a western-style diet is associated with hand-held computerized systems for temporal tracking of appetite ratings, Br. J.
poorer inhibition of wanting for palatable snack foods when sated. Journal of Nutr. 84 (2000) 405–415.
Experimental Psychology: Animal Learning and, Cognition 42 (2016) 415–428. [35] H. Tibboel, J. De Houwer, A. Spruyt, M. Field, E. Kemps, G. Crombez, Testing the
[2] Berridge, K. & O'Doherty, J. (2014). From experienced utility to decision utility. In, validity of implicit measures of wanting and liking, J. Behav. Ther. Exp. Psychiatry
Glimcher, P. & Fehr, E. (Eds), Neuroeconomics (pp.335–351). 2ND Edition: San 42 (2011) 284–292.
Diego: AP. [36] H. Tibboel, J. De Houwer, B. Van Bockstaele, Implicit measures of “wanting” and
[3] K. Berridge, T. Robinson, The mind of an addicted brain: neural sensitization of “liking” in humans, Neurosci. Biobehav. Rev. 57 (2015) 350–364.
wanting versus liking, Curr. Dir. Psychol. Sci. 4 (1995) 71–76. [37] O. Wooley, S. Wooley, Relationship of salivation in humans to deprivation, in-
[4] J. Born, S. Lemmens, M. Martens, E. Formisano, R. Goebel, M. Westerp-Plantenga, hibition and the encephalization of hunger, Appetite 2 (1981) 331–350.
Differences between liking and wanting signals in the human brain and relations [38] C. Wyvell, K. Berridge, Intra-accumbens amphetamine increases conditioned in-
with cognitive dietary restraint and BMI, Am. J. Clin. Nutr. 94 (2011) 392–403. centive salience of sucrose reward: enhancement of reward “wanting” without
[5] K. Burger, M. Cornier, J. Ingebrigsten, S. Johnson, Assessing food appeal and desire enhanced “liking” or response reinforcement, J. Neurosci. 20 (2000) 8122–8130.
to eat: the effects of portion size and energy density, Int. J. Behav. Nutr. Phys. Act. 8 [39] O. Van den Bergh, M. Walentynowicz, Accuracy and bias in retrospective symptom
(2011) 101. reporting, Current Opinion in Psychiatry 29 (2016) 302–308.
[6] M. Cabanac, Physiological role of pleasure, Science 173 (1971) 1103–1107.
[7] X. Dai, C. Brendl, D. Ariely, Wanting, liking and preference construction, Emotion