Science - USA (2021-11-12)

P=0.03andrGPi:Pearson’srdifference =
0.79;P= 0.007; Fig. 2, A to F, and supple-
mentary text). The difference was not sig-
nificant for the lCau (Pearson’srdifference =

• 0.08;P= 0.64).
  Overall, these findings establish that tool
  use and syntax rely on neural activity within
  common anatomical territories in the BG. The
  activity independently elicited by the two tasks
  displays similar spatial distribution, which is
  consistent with common neural resources. It
  has been documented that when two func-
  tions share neural resources and cognitive pro-

cesses, learning transfer occurs ( 41 ): Training

a specific ability can therefore benefit an un-

trained one ( 41 , 42 ). Therefore, we predicted

cross-domain learning transfer between tool

use and syntactic skills in language.

Learning transfer from tool use to syntactic

skills in language

In Experiment 2, we tested whether tool-use

training improves syntactic skills in language.

We used the same syntactic task as in Experi-

ment 1 to measure syntactic skills in 26 naïve,

healthy adults before and after tool-use train-

ing (Fig. 3A). The specificity of tool use was

controlled for by testing a distinct group (n=

26) undergoing an identical training regime

but with the free hand. A third passive control

group (n= 26) was also included to quantify

potential test-retest effects. These participants

were assessed in the same syntactic task be-

fore and after watching natural documentary

videos for an equivalent amount of time as the

two active groups engaged in motor training.

The three groups were comparable in terms of

relevant sociodemographic characteristics (see

the supplementary text).

Thibaultet al.,Science 374 , eabe0874 (2021) 12 November 2021 4of14

ABC

DEF

Fig. 2. Experiment 1: Functional link between tool-use planning and syntax
in the BG.(AtoC) Distribution, after 10,000 permutations, of differences of
PearsonÕsrcomputed for the correlations between patterns of brain activity for
object-relative clauses and those for motor planning of tool use and free hand
separately for the lCau (A), lGPi (B), and rGPi (C). Larger positive differences
indicate stronger similarity between tool-use planning and object-relative

processing compared with free-hand planning and object-relative processing. The

black line indicates the observed difference, and the red dotted line indicates

theP-value threshold set at 0.05. (DtoF) Spatial distribution of neural activity

for tool-use planning, object-relative planning, and free-hand planning in the

BG. Each single colored square represents a single voxel for the lCau (D), lGPi

(E), and rGPi (F).

RESEARCH | RESEARCH ARTICLE