Content uploaded by Helena Matute
Author content
All content in this area was uploaded by Helena Matute
Content may be subject to copyright.
main two terms of the to-be-learned conjunction, from con-
ditioning or quasi-conditioning paradigms of learning, primarily
those that have been frequently characterized as formed by
an associative “link.” Assuming that “learning” refers to the
acquisition of new knowledge, in this commentary I show
(necessarily briefly) that the notion that human associative
learning is neither automatic, nor necessarily unconscious, has
a venerable, nearly century-old history, missing from the
target article. Furthermore, propositional structures constitute
just one part of organization theories (see Mandler 2007, for a
more extended history).
The opposition to unconscious, automatic associative pro-
cesses started in modern times with the work of G. E. Mu¨ ller
(e.g., Mu¨ ller 1911) and proceeded rapidly with the development
of Gestalt theory. Following the work of Wertheimer (1921),
Duncker (1926), and Katona (1940), the time was ripe for a
full-scale assault on the mechanisms of associative memory.
The initial arguments were primarily presented by Asch (1962;
1969) and Asch and Ebenholtz (1962), and generated specific
demonstrations of human associative learning by Bower (1970),
Bower and Bryant (1991), Mandler and Mandler (1964),
Mandler (1968; 1979a; 1979b), and Murdock (1966). In
Mandler (1979b), I suggested three possible structures account-
ing for human associative phenomena: coordination (holistic,
unitary organizations), subordination (hierarchical organiz-
ations), and pro-ordination (sequential organization). The last
is most like the propositional structure proposed in the target
article – Afollowed by B.
Relevant to the target argument, my colleagues and I have
tested human associative learning and demonstrated that holistic
structures characterize the storage of verbal associations.
Mandler et al. (1981) showed that in verbal human associative
learning (sometimes known as paired-associates), “associations”
are stored not as “links,” but by combining the two terms in a
single holistic unit. Tests of free recall, cued-recall, and recog-
nition supported that conclusion.
Propositions about and tests of organizational theory describe
the structure of human semantics – the mental organization of
meaningful knowledge and experience. Organization defines
the structure of memory. It is obvious that propositional struc-
tures depend on retrievals from memory, and, albeit without
any detailed discussion of memory, Mitchell et al. too assert
the centrality of memorial functions, when in section 3.1 (para.
1) they state that the encoding of an associative hypothesis in
memory constitutes learning. Organization theory has generally
avoided any distinction between learning and memory. The
history of the organizational approach discussed the organization
of mental contents, which can be seen as “learned” when estab-
lished and retrieved once the organizational structure is estab-
lished. Consistent with such an approach, Mitchell et al. also
note that subsequent to a bell-food pairing, a bell can retrieve
memories of previous pairings. More generally, it may not be
initially obvious which of the possible structures applies to a par-
ticular learning experiment or paradigm. At present it is not
obvious which experimental or experiential situations give rise
to one organization or another. The target article seems to
claim that all encodings are propositional; in contrast, we have
shown that some are holistic and unitary. Specific experimental
procedures and probing and testing procedures need to be devel-
oped in order to determine which particular structures eventuate
from a specific “learning” situation.
Finally, it does not seem obvious that “we have been heading
... towards a propositional approach to all learning” (sect. 7.1,
para. 3). The holistic encoding of word pairs or the hierarchical
organization of some lists argues against a single model of under-
lying structures. A general organizational approach has asserted
for some time that learning is indeed “not separate from other
cognitive processes” (sect. 8, para. 1). Organization theory has
made it possible to see the connectedness of these various func-
tions and processes.
The Proust effect and the evolution of a dual
learning system
doi:10.1017/S0140525X09001046
Helena Matute and Miguel A. Vadillo
Department of Psychology, Deusto University, 48080 Bilbao, Spain.
matute@fice.deusto.es
http://paginaspersonales.deusto.es/matute
mvadillo@fice.deusto.es
http://paginaspersonales.deusto.es/mvadillo
Abstract: Proust’s madeleine illustrates the automatic nature of
associative learning. Although we agree with Mitchell et al. that no
compelling scientific proof for this effect has yet been reported in
humans, evolutionary constraints suggest that it should not be
discarded: There is no reason by which natural selection should favor
individuals who lose a fast and automatic survival tool.
And soon, mechanically, weary after a dull day with the prospect of a
depressing morrow, I raised to my lips a spoonful of the tea in which
I had soaked a morsel of the cake. No sooner had the warm liquid,
and the crumbs with it, touched my palate than a shudder ran
through my whole body, and I stopped, intent upon the extraordinary
changes that were taking place ... at once the vicissitudes of life had
become indifferent to me, its disasters innocuous, its brevity illusory.
— Marcel Proust (1913/1922), Remembrance of Things Past
The episode of the madeleine in the Proust work cited above is
one of the most famous passages of universal literature of all
times. Not only is it beautifully written, but the passage also
describes an experience that is so personal and so ubiquitous in
human nature that any psychologist, from Freud to Pavlov,
would love to explain it. We will refer to it as the “Proust
effect.” To our knowledge, it is the best possible description of
associative learning.
The beauty of the target article by Mitchell et al. is that it tries
to understand the Proust effect in its entirety, not just a part of it.
As such, the article is ambitious, important, and timely. It makes
us rethink all the established assumptions about learning. Con-
trary to all intuitions, Mitchell et al. (almost) convince us that
(a) there must be only one learning process, and (b) this
unique process must be propositional in nature.
The standard explanation for associative learning is the link
approach. Because the narrator in Proust’s novel had associated
the madeleines with all the happiness of childhood (even though
he was not aware of this fact), then tasting one of those cakes
now, after so many years, brought back the enormous happiness
and all the good feelings from childhood. Thus, the Proust effect
reflects a simple, automatic link that was created during child-
hood and is now expressed, also without effort or knowledge of
the contingencies, in the form of a conditioned response (CR).
According to the link proponents, there was no propositional
learning here, no consciousness of the contingencies while the
association was acquired; not even now that it is expressed.
Indeed, it will still take the narrator many pages and a consider-
able amount of thinking and elaborated reasoning to discover
why the madeleine was producing the CR.
But the link approach is not as simple as it seems, and Mitchell
et al. are correct in highlighting this point: The link approach pre-
supposes a dual (and complex) system. Automatic links need to
be complemented with some more-elaborated, rational, and
time-consuming forms of learning. This complex learning is at
work, for instance, after the CR has occurred and the narrator
begins to consciously think about it and tries to identify its
cause. Even the most enthusiastic proponents of low-level mech-
anisms have to admit that people are obviously capable of other
forms of learning and reasoning.
What Mitchell et al. suggest is that, if we all agree that prop-
ositional learning is needed, why should we maintain a belief in
automatic links? Couldn’t we assume just a propositional learning
Commentary/Mitchell et al.: The propositional nature of human associative learning
BEHAVIORAL AND BRAIN SCIENCES (2009) 32:2 215
process that could account for both the automatic-like and the
more complex processes? Are there any experiments that can
only be explained by the link mechanism? That there are data
to support that propositional learning exists is unquestionable,
and the authors make an excellent case of it. That many of the
results that have traditionally been explained using the link
approach can also be explained by the propositional account is
also clear in their target article. Moreover, it is well established
today that there are very few experiments that can be explained
solely by the link approach (Lovibond & Shanks 2002; Shanks &
St. John 1994). What Mitchell et al. are showing is that both the
dual and the propositional account can explain the majority of
the available evidence. Scientific parsimony becomes then the
central argument: If a single process can explain it all, why
should science maintain two?
But the argument of scientific parsimony should be confronted
against that of natural selection. A simple, low-level process is
vital for survival because, by definition, it can do all those
things the complex process cannot do: it responds quickly, auto-
matically, and without consciousness or effort to the demands of
the environment. Even under high pressure, it provides a fast
tool for survival. Its loss would be too costly.
As Mitchell et al. note, natural selection has produced a conti-
nuum of complexity in the different species. At one end of this
continuum, we find very simple species which have just the
link system and no cognition. At the other end, we find
the human species, which, according to Mitchell et al., has only
the propositional system. If so, Mitchell et al. need to explain
why humans (and other evolved animals) should have lost their
primitive link system while developing the propositional one.
There is no clear evolutionary advantage in losing a fast and auto-
matic tool.
Indeed, there is a growing body of evidence suggesting that
learning is actually caused by a multiplicity of different mechan-
isms and that the insistence of traditional learning theory in a
unique, general-purpose learning system was simply a mistake
(Gallistel 2000; Tooby & Cosmides 1992; 2005). If natural selec-
tion has encouraged flexibility and adaptability, having many
different forms of learning must have been favored through the
course of evolution.
In sum, Mitchell et al. need to explain not only why conscious-
ness becomes so difficult in the Proust effect, but also what sur-
vival advantages a species that extinguishes the link system
should have. If all the evidence for the automatic mechanism
would come from novels and intuitions, Mitchell et al. would
be right that science should ignore it. But we have shown good
reasons to believe that the automatic mechanism must still be
present in humans. Perhaps the problem is that the Proust
effect has always been taken for granted and proofs have not
been searched in the right places.
ACKNOWLEDGMENT
The authors of this commentary were supported by MICINN (Spain)
Grant SEJ2007-63691.
Both rules and associations are required to
predict human behaviour
doi:10.1017/S0140525X09001058
I. P. L. McLaren
School of Psychology, University of Exeter, Exeter, EX4 4QG, United Kingdom.
i.p.l.mclaren@exeter.ac.uk
Abstract: I argue that the dual-process account of human learning
rejected by Mitchell et al. in the target article is informative and
predictive with respect to human behaviour in a way that the authors’
purely propositional account is not. Experiments that reveal different
patterns of results under conditions that favour either associative or
rule-based performance are the way forward.
In this target article, Mitchell et al. argue for a propositional
account of human learning, rather than a dual-process model
that allows for propositional and associative (what they call the
“link model”) processes to operate concurrently. The issue at
hand, then, is whether we need to postulate associative processes
in addition to propositional ones; the converse argument,
whether we need to postulate propositional processes in addition
to associative processes, can be left for another time. But let me
be quite clear: I am of the view that we need to appeal to both if
we are to understand learning in humans.
The approach taken in this commentary is to point out differ-
ences in learning and performance under conditions that should
favour either propositional or associative learning. Mitchell et al.
consider a number of these cases, but perhaps do not do them
justice. I take as my first example their review of the Le Pelley
et al. (2005a) demonstration of unblocking in humans. In these
experiments, it was demonstrated that a design such as
A2.O
1
followed by O
2
jAB2. O
1
followed by O
3
revealed
that learning to Bwas greater for outcome 1 (O
1
) than in a con-
ventional blocking design where the second phase had the com-
pound followed by O
1
then O
2
. This finding was predicted on the
basis of Mackintosh’s (1975) associative theory of learning, which
has received experimental support in animals other than human.
To dismiss it by saying that it is possible that in a complex design
the human participants had forgotten earlier trials and knew
something had changed but were not sure whether it was O
1
or
O
2
, ignores this background. As an explanation of the phenom-
enon, it is terrifically weak. We are expected to allow that prop-
ositional learning and an automatic memory (that is definitely
not associative?) are both imperfect, and so people make mis-
takes, which just happen to be the ones that associative theories
predict.
This does sound rather implausible, and it is, even though the
authors reassure us that it can be tested. Their proposal is to
make the outcomes more distinctive, thus reducing any con-
fusion between them, and so the effect (unblocking) should go
away. In fact, if the outcomes were made that distinct from one
another, the same associative theory that predicted the original
result would now predict that the effect would go away as well,
as the alpha change that leads to unblocking is to some extent
reinforcer-specific in this model. This result has also been
found in humans in another experiment by Le Pelley and col-
leagues (Le Pelley et al. 2005b), in which changing outcomes
from those that are generally “nice” to those that are generally
“nasty” (and vice versa) prevented alpha effects that were gener-
ated by manipulating the predictiveness of certain cues during
training. So we are left with a “test” of their account that fails
to distinguish between it and the very associative theory that
motivated the experiment in the first place. Not much of a test!
Mitchell et al. also fail to take into consideration a number of
other studies that demonstrate a different pattern of results when
learning is dominated by either rule-based (hence propositional)
or associative processes. People show a peak shift, like pigeons,
when they are tested on a dimension after relatively little experi-
ence with it, and when they are unable to verbalise any rule that
captures the discrimination (Jones & McLaren 1999; and see
Livesey & McLaren, forthcoming). This pattern of responding
changes (to a monotonic function across the dimension) after
extensive experience with the stimuli and when people can ver-
balise the correct rule. In the spirit of the target article, I
would expect the response to be that this does not demonstrate
associative learning, but instead, incorrect rule induction or
imperfect application of a rule in some way. If this characteris-
ation of Mitchell et al.’s position is right, then it is impossible
to defend against. There will always, with sufficient ingenuity,
be some incorrect or imperfect rule that can be appealed to
Commentary/Mitchell et al.: The propositional nature of human associative learning
216 BEHAVIORAL AND BRAIN SCIENCES (2009) 32:2
