There is an ongoing debate in the literature about whether the present global warming is increasing local and global temperature variability. The central methodological issues of this debate relate to the proper treatment of normalised temperature anomalies and trends in the studied time series which may be difficult to separate from time-evolving fluctuations. Some argue that temperature variability is indeed increasing globally, whereas others conclude it is decreasing or remains practically unchanged. Meanwhile, a consensus appears to emerge that local variability in certain regions (e.g. Western Europe and North America) has indeed been increasing in the past 40 years. Here we investigate the nature of connections between external forcing and climate variability conceptually by using a laboratory-scale minimal model of mid-latitude atmospheric thermal convection subject to continuously decreasing 'equator-to-pole' temperature contrast ΔT, mimicking climate change. The analysis of temperature records from an ensemble of experimental runs ('realisations') all driven by identical time-dependent external forcing reveals that the collective variability of the ensemble and that of individual realisations may be markedly different - a property to be considered when interpreting climate records.