Short- and Long-Term Implicit Memory in Aging and Alzheimer's Disease

ABSTRACT

Implicit memory processes were investigated via picture naming in healthy young and older adults and in persons with mild Alzheimer's disease (AD). Repetition priming in picture-naming was intact in all groups over the course of a short retention interval (seconds), and only the AD group revealed a deficit over a longer interval (72 hours). In addition, the AD group showed impaired procedural memory, with no benefit of practice on picture-naming. Impaired long-term priming was related to severity of AD. Both theoretical and methodological implications are discussed.

ACKNOWLEDGMENTS

Portions of these data were presented at the meeting of the International Neuropsychological Society, Houston in February 1984, and at the meeting of the Psychonomic Society, Chicago in November 1988.

We thank Drs. Albert Heyman and Donald Schmechel for making the AD patients available. Audrey Norman and the medical personnel on the Rankin Ward provided valuable assistance. Jennifer Rogers supplied bodacious bibliographic support. Data collection was supported by NIA grant #AG00029 to the Center for the Study of Aging and Human Development, Duke University Medical Center. Support for bibliographic research, data analysis, and write-up was provided by grants from Dedman College, NIA #AG07854, and the Foley Family Foundation to David Mitchell, and NIA grant #5P50AG005144 to the Sanders-Brown ADRC.

Notes

*1–8 intervening lists; our estimate of interval.

**collapsed over words presented 1–4 times.

¹These individuals also participated in an episodic memory study (Mitchell, Hunt, & Schmitt, 1986).

²The repeated picture was the 5th slide after the first occurrence, i.e., four other pictures intervened. The low-end estimate for this interval is approximately 8–10 sec (≤ 1 sec for naming, ≤ 1 sec ISI), and the high-end estimate is approximately 20 sec (≤ 2 sec for naming, ≤ 3 sec ISI for experimenter to record a nondominant response). Thus, the five-lag items were generally within the accepted range of short-term memory life span. Although these intervals were successively longer for the older and AD participants on the average, these minor variations in seconds appear to have had no impact on short-term priming. The five-item lag has been used previously (Mitchell, 1989). A shorter lag (e.g., 0) seems to produce some sort of a startle response, an effect not of the same variety as the phenomenon of repetition priming.

³To measure latency, we hooked a voice key to a computer to give us millisecond accuracy between tone and voice onset (and maximum deflection/voice volume). In addition, we visually presented the digitized signal of the voice onset and tone for comparison with the computer analog to digital signals as marked by the voice key. This allowed us to verify that the timing was actually tone-word onset (first syllable) and also ignore background noise. In addition, we ran several hundred blank slides to adjust for the “dropping” of the slide (visual onset of picture) using a light meter hooked into the computer (tone-to-light meter deflection) and tone-to-slide drop (sound by voice key).

⁴The data from healthy older adults resulted in the addition of a few nondominant names not found in the Snodgrass and Vanderwart (1980) norms. For example, they listed no nondominant names for mushroom, but “toadstool” was not an uncommon name for this item among the elderly. The norms (Mitchell & Brown, 1993) are available from David Mitchell.

^aTrials = mean number of pairs of usable response latency trials of 16 possible pairs per condition.

*These particular maximums identified as outliers by both parametric and nonparametric criteria (see text).

¹Primers = the number of participants (out of 12 per group) who exhibited positive priming (i.e., faster latencies on the second naming).

²Outliers = number of data points excluded when both parametric and nonparametric criteria were met (see text); 1st number = msec priming/2nd number = proportional priming.

*Priming was not statistically reliable for these means. All other priming means were significantly greater than zero (one sample t-tests, ps < .05).