relationships’’. While we agree with this statement, we emphasize
below that the sole description of behavioral contexts is not
necessarily a step forward towards the appraisal of causal
environmental effects on behavior.
The authors reported that, after adjustment, the odds of high
physical activity intensity were higher in GPS locations with
parks, schools, and high population density, and lower in GPS
locations with more roads and food outlets. The descriptive
nature of these findings is illustrated, for example, by the
argument that the lower odds of intense physical activity near
food outlets ‘‘may be capturing sedentary behavior, when participants
visit malls with outdoor areas, or restaurants with outdoor
seating’’. These findings simply suggest that people are by
essence less physically active in specific places (e.g., restaurants,
movie theaters, etc.) than in others (e.g., parks).
The third reviewed study (Almanza et al., 2012) relied on GPS
and accelerometer data (30 s intervals/epochs) collected for 7
days for 208 children aged 8–14 years from The Preserve smart
growth community in California (USA) and six conventional
communities situated nearby. Interestingly, the study was
designed to rule out selective residential migration biases by
comparing families who moved to the smart growth community
with families who initially considered moving there but did not.
Analytically, the authors compared the contemporaneous
momentary analytical design used in the Lachowycz and Rodriguez
studies (epoch-level analyses) with a more conventional
individual-level analysis.
Contemporaneous momentary analyses revealed a positive
relationship between greenness at the GPS point and the likelihood
of moderate-to-vigorous physical activity. In individuallevel
analyses, greenness exposure in the residential neighborhood
was defined in two ways: (i) average greenness in the 500 m
buffer around the residence and (ii) cumulated time of exposure
to greenness at all the GPS points recorded in the residential
neighborhood. The association between greenness and physical
activity identified in the momentary analysis was retrieved only
with the second version of the individual-level greenness exposure
variable.
Because of their ‘‘spatially-explicit’’ design (considerable number
of locations examined for each participant), contemporaneous
momentary analyses were described by the authors as increasing
the power to detect associations compared to individual-level
analyses. Whether true or not, such simple epoch-level analyses
which assess the spatial milieu around individuals at each
observation are useful to describe behavioral contexts, but they
may be inadequate to assess environmental effects on behavior.
For example, such simple contemporaneous momentary analyses
are unable to demonstrate that an improved spatial accessibility
to greenness causally increases physical activity; they simply
highlight that green spaces are a more common place for physical
activity than many other places such as railway stations or
shopping areas.
The individual-level greenness exposure variable defined in
500 m radius buffers around the residence was qualified as ‘‘coarser’’
than the individual-level variable based on the aggregation of
greenness exposure at GPS activity locations. Again, in our view,
the latter variable is not only more accurate, its meaning is also
qualitatively different: whereas the former variable reflects potential
access to green spaces from the residence, the latter captures the
actual patterns of use of green spaces in local daily trajectories. As
discussed below, such difference has major implications for the
interpretation of the associations estimated between the environmental
variables and the behavioral or health outcomes.
The standard contemporaneous momentary design was
described above as providing descriptive information on behavioral
contexts. However, the momentary analysis by Almanza et al.