An fMRI study of relational complexity effects in a kanji pair list recognition task

Steven Phillips, Kazuhisa Niki, Jing Luo
Electrotechnical Laboratory
Chinese Academy of Sciences

Abstract
Two general theories of cognitive capacity are based on "number of elements"
stored in working memory (Miller, 1956) versus the "complexity of relations"
between them (Halford, et al, 1998). Here, we contravaried number of unique
elements (six versus four) with relational arity (unary versus binary) in
a Japanese kanji recognition task. fMRI analysis revealed greater prefrontal
(executive control) and parietal (visual buffer) activity in the binary-four
element condition than in the unary-six element condition, lending further
support for relational complexity theory.

Design:

Six Japanese university students were tested in a "kanji pair recognition"
experiment. Each subject was given 60 trials, consisting of an encoding phase
of three kanji pairs (presented one per time interval); a probing phase of one
pair (either from the list, or a repaired left-right combination); and a response
phase indicating whether the probe was (not) a pair in the prior list. Pairs were
constructed, and screened by a native Japanese speaker, so as to be meaningless,
minimizing semantic and phonetic associations; and roughly balanced by stroke count.
We used a 3 (baseline,unary,binary) x 2 (target,distractor) design, with list
conditions AA,BB,CC(baseline); AB,CD,EF(unary); AB,CB,AD(binary). Each left
or right kanji uniquely identified a pair in the unary condition (index length 1);
but not in the binary condition (index length 2). Conditions and pairs were randomly
ordered and counterbalanced, but most target probes matched the middle pair, to
avoid primacy and recency effects.

Analysis:

Scanning was performed on a 3.0-T MRI Scanner (GE 3T Signa) with EPI capability.
19 axial slices (5.5 mm thick, 1 mm spaces, interleaved) covering the entire brain.
A T2* weighted gradient echo EPI was employed. The imaging parameters were TR=2 sec,
TE=32 ms, FA = 70 degrees, FOV=20x20 (64x64 mesh). Images were pre-processed
(timeslice adjusted, realigned, normalized and smoothed) by SPM99. Data were
estimated to establish a fixed model in which 14 event types were defined:
6 (trial) x 2 (phases: encode, probe) + 2 (errors: target, distractor).
Initial analysis is reported for unary and binary conditions, collapsed over
target and distractor trials (correct responses only). Event types were:
binary-encode(BE), binary-probe(BP), unary-encode(UE), unary-probe(UP).
The threshold was set at p < 0.001, uncorrected.

Results:

Greater activity was observed in prefrontal (executive control) and parietal
(visual buffer) lobes in binary than unary conditions for both encoding and
probing phases. Prefrontal activity was more anterior and dorsal (area 8)
in the unary condition in the encoding phase, but this was only
marginally significant at p < 0.063, corrected (Figure).
Mean subject response errors were 6.2 (unary) and 2.8 (binary), but
this difference was not significant (p > 0.2, one-sided t-test).

Discussion:

Increased activity in areas usually regarded as executive control and visual buffer
components of working memory (binary condition) supports relational complexity theory,
but not a "number of elements" theory of capacity, where unique elements were fewer.
Contra Cowan (2001), these results suggest capacity is not reducible to number of
task elements, but must include the complexity of relations between them
(Halford, et al, 1998). However, more work is needed to clarify the role of
area 8 in the unary condition.

References:

Cowan (2001). The magical number 4 in short-term memory.
Behavioral and Brain Sciences.

Halford, Wilson and Phillips (1998). Processing capacity defined by relational complexity.
Behavioral and Brain Sciences.

Miller (1956). The magical number seven, plus or minus two.
Psychological Review.