Global loudness change is a post-stimulus retrospective judgement that measures listeners' overall impressions of loudness change in response to stimuli with continuous increases (up-ramps) and decreases (down-ramps) of acoustic intensity that are otherwise acoustically identical. Past results indicate that global loudness change is significantly greater in response to up-ramps relative to down-ramps for tonal stimuli (e.g., vowel) but not white-noise. An adaptive perceptual bias for up-ramp tonal stimuli has been proposed as a functional ecological explanation. However, global loudness change may also be influenced by stimulus duration and an end-level recency-in-memory mechanism that biases retrospective global judgements on a ramp's end-level intensity, rather than its entire magnitude of intensity change. The present within-subjects experiment (N = 34) was designed to systematically investigate the effects of intensity, spectrum, and duration on global loudness change when end-level recency is controlled. Up-ramps and down-ramps were embedded within two spectral conditions (tonal vowel /@/ and white-noise) and presented over three durations (1.8 s, 3.6 s, 7.2 s) and two regions of intensity change (45-65 dB SPL, 65-85 dB SPL). End-level recency response bias was controlled through the use of balanced end-level comparisons between 45-65 dB SPL up-ramps and 85-65 dB SPL down-ramps that both converged on 65 dB SPL. Overall, global loudness change was significantly greater in response to vowel and white-noise up-ramps, relative to their corresponding down-ramps. However, with end-level recency controlled, global loudness change was significantly greater for up-ramps relative to down-ramps in 3.6 s and 7.2 s vowel conditions only. This was facilitated by an up-ramp-specific effect of duration, where the magnitude of global loudness change increased as vowel up-ramp duration increased from 3.6 s to 7.2 s. The findings are discussed in the context of psychoacoustics and ecological acoustics.