information. As I discussed in the fourth section, the record-keeping theory
suggests that interference is primarily due to the distracting effects of other
memory records, which increase in number as the amount of information to
be remembered increases. But the record-keeping theory implies a paradox:
Adults should show a gradual decline in their memory as they learn more
about various topics. Experts should have especially poor memory in their
domains of expertise. Yet neither of these propositions is true.
The constructionist theory predicts interference when no distinctive patterns
enable the person to reconstruct information, as is likely to happen in list
memorization experiments. Because the constructionist theory claims that no
memory records are kept, an adult’s memory remains stable over time. Because
experts learn to find patterns in and to develop reconstruction strategies for
their domain of expertise, experts have a good memory for that domain. The
constructionist theory correctly predicts that interference usually observed in
list learning experiments is eliminated if the memorized facts can be integrated
by a common theme, or if the memory test requires people to remember the
patterns rather than the details contained within the memorized material.
Although constructionist accounts of memory are currently influential (see
Schacter, 1996), some cognitive psychologists continue to support record-
keeping theories (see Hall, 1990). One might argue that, with suitable mod-
ifications, the record-keeping theory can explain the data I claimed support the
constructionist theory. For example, a record-keeping theory could include a
pattern recognition system that either stores descriptions of patterns or examines
memory records to find patterns in events. Consequently, patterns of experi-
ences would be readily remembered. A record-keeping theory could posit that
reconstruction strategies are used when a sought-after memory record is not
located.
It is true that such modifications would make the record-keeping theory
work more like real human memory. Note, though, that the proposed mod-
ifications have the effect of making the record-keeping theory more like the
constructionist theory. Furthermore, the modifications are not intrinsic to, or a
natural consequence of, the central idea that memory is a matter of storing
records of experiences. There is nothing about putting a record of an experience
someplace in a storage bin that inevitably leads to extracting a pattern. There is
nothing about reading memory records that leads to making plausible guesses
about what happened in the past. These modifications are just tacked on, be-
cause without them the system does not resemble human memory. To put it
another way, the record-keeping theory so modified lacks theoretical elegance.
In contrast, consider that the central idea of the constructionist theory, that
the cognitive systems change the strength of their connections in response to
events, does lead naturally to how human memory actually works. Remem-
bering patterns, but not details, is a natural consequence of such a system, be-
cause the invariants in experiences strengthen already existing connections. No
pattern recognition system has to be added on. Reconstruction happens be-
cause no records of past experiences are ever ‘‘read’’ or ‘‘reexperienced’’ ; rather,
past events must be inferred from the current state of connections. And a con-
structionist theory of memory more closely reflects what is known about the
neurophysiology of learning and remembering.
Memory 351